CS236781: Deep Learning on Computational Accelerators¶

Homework Assignment 2¶

Faculty of Computer Science, Technion.

Submitted by:

# Name Id email
Student 1 Elie Nedjar 336140116 elie.nedjar@campus.technion.ac.il
Student 2 David El Beze 931177034 edavid@campus.technion.ac.il

Introduction¶

In this assignment we'll create a from-scratch implementation of two fundemental deep learning concepts: the backpropagation algorithm and stochastic gradient descent-based optimizers. In addition, will create a general-purpose multilayer perceptron, the core building block of deep neural networks. We'll visualize decision boundaries and ROC curves in the context of binary classification. Following that we will focus on convolutional networks with residual blocks. We'll use PyTorch to create our own network architectures and train them using GPUs on the course servers, and we'll conduct architecture experiments to determine the the effects of different architectural decisions on the performance of deep networks.

General Guidelines¶

  • Please read the getting started page on the course website. It explains how to setup, run and submit the assignment.
  • Please read the course servers usage guide. It explains how to use and run your code on the course servers to benefit from training with GPUs.
  • The text and code cells in these notebooks are intended to guide you through the assignment and help you verify your solutions. The notebooks do not need to be edited at all (unless you wish to play around). The only exception is to fill your name(s) in the above cell before submission. Please do not remove sections or change the order of any cells.
  • All your code (and even answers to questions) should be written in the files within the python package corresponding the assignment number (hw1, hw2, etc). You can of course use any editor or IDE to work on these files.

Contents¶

  • Part 1: Backpropagation
  • Part 2: Optimization and Training:
  • Part 3: Binary Classification with Multilayer Perceptrons
  • Part 4: Convolutional Neural Networks:
  • Part 5: Convolutional Architecture Experiments
$$ \newcommand{\mat}[1]{\boldsymbol {#1}} \newcommand{\mattr}[1]{\boldsymbol {#1}^\top} \newcommand{\matinv}[1]{\boldsymbol {#1}^{-1}} \newcommand{\vec}[1]{\boldsymbol {#1}} \newcommand{\vectr}[1]{\boldsymbol {#1}^\top} \newcommand{\rvar}[1]{\mathrm {#1}} \newcommand{\rvec}[1]{\boldsymbol{\mathrm{#1}}} \newcommand{\diag}{\mathop{\mathrm {diag}}} \newcommand{\set}[1]{\mathbb {#1}} \newcommand{\norm}[1]{\left\lVert#1\right\rVert} \newcommand{\pderiv}[2]{\frac{\partial #1}{\partial #2}} \newcommand{\bb}[1]{\boldsymbol{#1}} $$

Part 1: Backpropagation¶

In this part we will learn about backpropagation and automatic differentiation. We'll implement both of these concepts from scratch and compare our implementation to PyTorch's built in implementation (autograd).

In [2]:
import torch
import unittest

%load_ext autoreload
%autoreload 2

test = unittest.TestCase()
The autoreload extension is already loaded. To reload it, use:
  %reload_ext autoreload

The backpropagation algorithm is at the core of training deep models. To state the problem we'll tackle in this notebook, imagine we have an L-layer MLP model, defined as $$ \hat{\vec{y}^i} = \vec{y}_L^i= \varphi_L \left( \mat{W}L \varphi{L-1} \left( \cdots \varphi_1 \left( \mat{W}_1 \vec{x}^i + \vec{b}_1 \right) \cdots \right)

  • \vec{b}_L \right), $$

a pointwise loss function $\ell(\vec{y}, \hat{\vec{y}})$ and an empirical loss over our entire data set, $$ L(\vec{\theta}) = \frac{1}{N} \sum_{i=1}^{N} \ell(\vec{y}^i, \hat{\vec{y}^i}) + R(\vec{\theta}) $$

where $\vec{\theta}$ is a vector containing all network parameters, e.g. $\vec{\theta} = \left[ \mat{W}_{1,:}, \vec{b}_1, \dots, \mat{W}_{L,:}, \vec{b}_L \right]$.

In order to train our model we would like to calculate the derivative (or gradient, in the multivariate case) of the loss with respect to each and every one of the parameters, i.e. $\pderiv{L}{\mat{W}_j}$ and $\pderiv{L}{\vec{b}_j}$ for all $j$. Since the gradient "points" to the direction of functional increase, the negative gradient is often used as a descent direction for descent-based optimization algorithms. In other words, iteratively updating each parameter proportianally to it's negetive gradient can lead to convergence to a local minimum of the loss function.

Calculus tells us that as long as we know the derivatives of all the functions "along the way" ($\varphi_i(\cdot),\ \ell(\cdot,\cdot),\ R(\cdot)$) we can use the chain rule to calculate the derivative of the loss with respect to any one of the parameter vectors. Note that if the loss $L(\vec{\theta})$ is scalar (which is usually the case), the gradient of a parameter will have the same shape as the parameter itself (matrix/vector/tensor of same dimensions).

For deep models that are a composition of many functions, calculating the gradient of each parameter by hand and implementing hard-coded gradient derivations quickly becomes infeasible. Additionally, such code makes models hard to change, since any change potentially requires re-derivation and re-implementation of the entire gradient function.

The backpropagation algorithm, which we saw in the lecture, provides us with a effective method of applying the chain rule recursively so that we can implement gradient calculations of arbitrarily deep or complex models.

We'll now implement backpropagation using a modular approach, which will allow us to chain many components layers together and get automatic gradient calculation of the output with respect to the input or any intermediate parameter.

To do this, we'll define a Layer class. Here's the API of this class:

In [3]:
import hw2.layers as layers
help(layers.Layer)
Help on class Layer in module hw2.layers:

class Layer(abc.ABC)
 |  A Layer is some computation element in a network architecture which
 |  supports automatic differentiation using forward and backward functions.
 |  
 |  Method resolution order:
 |      Layer
 |      abc.ABC
 |      builtins.object
 |  
 |  Methods defined here:
 |  
 |  __call__(self, *args, **kwargs)
 |      Call self as a function.
 |  
 |  __init__(self)
 |      Initialize self.  See help(type(self)) for accurate signature.
 |  
 |  __repr__(self)
 |      Return repr(self).
 |  
 |  backward(self, dout)
 |      Computes the backward pass of the layer, i.e. the gradient
 |      calculation of the final network output with respect to each of the
 |      parameters of the forward function.
 |      :param dout: The gradient of the network with respect to the
 |      output of this layer.
 |      :return: A tuple with the same number of elements as the parameters of
 |      the forward function. Each element will be the gradient of the
 |      network output with respect to that parameter.
 |  
 |  forward(self, *args, **kwargs)
 |      Computes the forward pass of the layer.
 |      :param args: The computation arguments (implementation specific).
 |      :return: The result of the computation.
 |  
 |  params(self)
 |      :return: Layer's trainable parameters and their gradients as a list
 |      of tuples, each tuple containing a tensor and it's corresponding
 |      gradient tensor.
 |  
 |  train(self, training_mode=True)
 |      Changes the mode of this layer between training and evaluation (test)
 |      mode. Some layers have different behaviour depending on mode.
 |      :param training_mode: True: set the model in training mode. False: set
 |      evaluation mode.
 |  
 |  ----------------------------------------------------------------------
 |  Data descriptors defined here:
 |  
 |  __dict__
 |      dictionary for instance variables (if defined)
 |  
 |  __weakref__
 |      list of weak references to the object (if defined)
 |  
 |  ----------------------------------------------------------------------
 |  Data and other attributes defined here:
 |  
 |  __abstractmethods__ = frozenset({'backward', 'forward', 'params'})

In other words, a Layer can be anything: a layer, an activation function, a loss function or generally any computation that we know how to derive a gradient for.

Each Layer must define a forward() function and a backward() function.

  • The forward() function performs the actual calculation/operation of the block and returns an output.
  • The backward() function computes the gradient of the input and parameters as a function of the gradient of the output, according to the chain rule.

Here's a diagram illustrating the above explanation:

Note that the diagram doesn't show that if the function is parametrized, i.e. $f(\vec{x},\vec{y})=f(\vec{x},\vec{y};\vec{w})$, there are also gradients to calculate for the parameters $\vec{w}$.

The forward pass is straightforward: just do the computation. To understand the backward pass, imagine that there's some "downstream" loss function $L(\vec{\theta})$ and magically somehow we are told the gradient of that loss with respect to the output $\vec{z}$ of our block, i.e. $\pderiv{{L}{\vec{z}}}$.

Now, since we know how to calculate the derivative of $f(\vec{x},\vec{y};\vec{w})$, it means we know how to calculate $\pderiv{\vec{z}}{\vec{x}}$, $\pderiv{\vec{z}}{\vec{y}}$ and $\pderiv{\vec{z}}{\vec{w}}$ . Thanks to the chain rule, this is all we need to calculate the gradients of the loss w.r.t. the input and parameters:

$$ \begin{align} \pderiv{L}{\vec{x}} &= \pderiv{L}{\vec{z}}\cdot \pderiv{\vec{z}}{\vec{x}}\\ \pderiv{L}{\vec{y}} &= \pderiv{L}{\vec{z}}\cdot \pderiv{\vec{z}}{\vec{y}}\\ \pderiv{L}{\vec{w}} &= \pderiv{L}{\vec{z}}\cdot \pderiv{\vec{z}}{\vec{w}} \end{align} $$

Comparison with PyTorch¶

PyTorch has the nn.Module base class, which may seem to be similar to our Layer since it also represents a computation element in a network. However PyTorch's nn.Modules don't compute the gradient directly, they only define the forward calculations. Instead, PyTorch has a more low-level API for defining a function and explicitly implementing it's forward() and backward(). See autograd.Function. When an operation is performed on a tensor, it creates a Function instance which performs the operation and stores any necessary information for calculating the gradient later on. Additionally, Functionss point to the other Function objects representing the operations performed earlier on the tensor. Thus, a graph (or DAG) of operations is created (this is not 100% exact, as the graph is actually composed of a different type of class which wraps the backward method, but it's accurate enough for our purposes).

A Tensor instance which was created by performing operations on one or more tensors with requires_grad=True, has a grad_fn property which is a Function instance representing the last operation performed to produce this tensor. This exposes the graph of Function instances, each with it's own backward() function. Therefore, in PyTorch the backward() function is called on the tensors, not the modules.

Our Layers are therefore a combination of the ideas in Module and Function and we'll implement them together, just to make things simpler. Our goal here is to create a "poor man's autograd": We'll use PyTorch tensors, but we'll calculate and store the gradients in our Layers (or return them). The gradients we'll calculate are of the entire block, not individual operations on tensors.

To test our implementation, we'll use PyTorch's autograd.

Note that of course this method of tracking gradients is much more limited than what PyTorch offers. However it allows us to implement the backpropagation algorithm very simply and really see how it works.

Let's set up some testing instrumentation:

In [4]:
from hw2.grad_compare import compare_layer_to_torch

def test_block_grad(block: layers.Layer, x, y=None, delta=1e-3):
    diffs = compare_layer_to_torch(block, x, y)
    
    # Assert diff values
    for diff in diffs:
        test.assertLess(diff, delta)

# Show the compare function
compare_layer_to_torch??

Notes:

  • After you complete your implementation, you should make sure to read and understand the compare_layer_to_torch() function. It will help you understand what PyTorch is doing.
  • The value of delta above is should not be needed. A correct implementation will give you a diff of exactly zero.

Layer Implementations¶

We'll now implement some Layers that will enable us to later build an MLP model of arbitrary depth, complete with automatic differentiation.

For each block, you'll first implement the forward() function. Then, you will calculate the derivative of the block by hand with respect to each of its input tensors and each of its parameter tensors (if any). Using your manually-calculated derivation, you can then implement the backward() function.

Notice that we have intermediate Jacobians that are potentially high dimensional tensors. For example in the expression $\pderiv{L}{\vec{w}} = \pderiv{L}{\vec{z}}\cdot \pderiv{\vec{z}}{\vec{w}}$, the term $\pderiv{\vec{z}}{\vec{w}}$ is a 4D Jacobian if both $\vec{z}$ and $\vec{w}$ are 2D matrices.

In order to implement the backpropagation algorithm efficiently, we need to implement every backward function without explicitly constructing this Jacobian. Instead, we're interested in directly calculating the vector-Jacobian product (VJP) $\pderiv{L}{\vec{z}}\cdot \pderiv{\vec{z}}{\vec{w}}$. In order to do this, you should try to figure out the gradient of the loss with respect to one element, e.g. $\pderiv{L}{\vec{w}_{1,1}}$ and extrapolate from there how to directly obtain the VJP.

Activation functions¶

(Leaky) ReLU¶

ReLU, or rectified linear unit is a very common activation function in deep learning architectures. In it's most standard form, as we'll implement here, it has no parameters.

We'll first implement the "leaky" version, defined as

$$ \mathrm{relu}(\vec{x}) = \max(\alpha\vec{x},\vec{x}), \ 0\leq\alpha<1 $$

This is similar to the ReLU activation we've seen in class, only that it has a small non-zero slope then it's input is negative. Note that it's not strictly differentiable, however it has sub-gradients, defined separately any positive-valued input and for negative-valued input.

TODO: Complete the implementation of the LeakyReLU class in the hw2/layers.py module.

In [5]:
N = 100
in_features = 200
num_classes = 10
eps = 1e-6
In [6]:
# Test LeakyReLU
alpha = 0.1
lrelu = layers.LeakyReLU(alpha=alpha)
x_test = torch.randn(N, in_features)

# Test forward pass
z = lrelu(x_test)
test.assertSequenceEqual(z.shape, x_test.shape)
test.assertTrue(torch.allclose(z, torch.nn.LeakyReLU(alpha)(x_test), atol=eps))

# Test backward pass
test_block_grad(lrelu, x_test)
Comparing gradients... 
input    diff=0.000

Now using the LeakyReLU, we can trivially define a regular ReLU block as a special case.

TODO: Complete the implementation of the ReLU class in the hw2/layers.py module.

In [7]:
# Test ReLU
relu = layers.ReLU()
x_test = torch.randn(N, in_features)

# Test forward pass
z = relu(x_test)
test.assertSequenceEqual(z.shape, x_test.shape)
test.assertTrue(torch.allclose(z, torch.relu(x_test), atol=eps))

# Test backward pass
test_block_grad(relu, x_test)
Comparing gradients... 
input    diff=0.000

Sigmoid¶

The sigmoid function $\sigma(x)$ is also sometimes used as an activation function. We have also seen it previously in the context of logistic regression.

The sigmoid function is defined as

$$ \sigma(\vec{x}) = \frac{1}{1+\exp(-\vec{x})}. $$
In [8]:
# Test Sigmoid
sigmoid = layers.Sigmoid()
x_test = torch.randn(N, in_features, in_features) # 3D input should work

# Test forward pass
z = sigmoid(x_test)
test.assertSequenceEqual(z.shape, x_test.shape)
test.assertTrue(torch.allclose(z, torch.sigmoid(x_test), atol=eps))

# Test backward pass
test_block_grad(sigmoid, x_test)
Comparing gradients... 
input    diff=0.000

Hyperbolic Tangent¶

The hyperbolic tangent function $\tanh(x)$ is a common activation function used when the output should be in the range [-1, 1].

The tanh function is defined as

$$ \tanh(\vec{x}) = \frac{\exp(x)-\exp(-x)}{\exp(x)+\exp(-\vec{x})}. $$
In [9]:
# Test TanH
tanh = layers.TanH()
x_test = torch.randn(N, in_features, in_features) # 3D input should work

# Test forward pass
z = tanh(x_test)
test.assertSequenceEqual(z.shape, x_test.shape)
test.assertTrue(torch.allclose(z, torch.tanh(x_test), atol=eps))

# Test backward pass
test_block_grad(tanh, x_test)
Comparing gradients... 
input    diff=0.000

Linear (fully connected) layer¶

First, we'll implement an affine transform layer, also known as a fully connected layer.

Given an input $\mat{X}$ the layer computes,

$$ \mat{Z} = \mat{X} \mattr{W} + \vec{b} ,~ \mat{X}\in\set{R}^{N\times D_{\mathrm{in}}},~ \mat{W}\in\set{R}^{D_{\mathrm{out}}\times D_{\mathrm{in}}},~ \vec{b}\in\set{R}^{D_{\mathrm{out}}}. $$

Notes:

  • We write it this way to follow the implementation conventions.
  • $N$ is the number of samples in the input (batch size). The input $\mat{X}$ will always be a tensor containing a batch dimension first.
  • Thanks to broadcasting, $\vec{b}$ can remain a vector even though the input $\mat{X}$ is a matrix.

TODO: Complete the implementation of the Linear class in the hw2/layers.py module.

In [10]:
# Test Linear
out_features = 1000
fc = layers.Linear(in_features, out_features)
x_test = torch.randn(N, in_features)

# Test forward pass
z = fc(x_test)
test.assertSequenceEqual(z.shape, [N, out_features])
torch_fc = torch.nn.Linear(in_features, out_features,bias=True)
torch_fc.weight = torch.nn.Parameter(fc.w)
torch_fc.bias = torch.nn.Parameter(fc.b)
test.assertTrue(torch.allclose(torch_fc(x_test), z, atol=eps))

# Test backward pass
test_block_grad(fc, x_test)

# Test second backward pass
x_test = torch.randn(N, in_features)
z = fc(x_test)
z = fc(x_test)
test_block_grad(fc, x_test)
Comparing gradients... 
input    diff=0.000
param#01 diff=0.000
param#02 diff=0.000
Comparing gradients... 
input    diff=0.000
param#01 diff=0.000
param#02 diff=0.000

Cross-Entropy Loss¶

As you know by know, cross-entropy is a common loss function for classification tasks. In class, we defined it as

$$\ell_{\mathrm{CE}}(\vec{y},\hat{\vec{y}}) = - {\vectr{y}} \log(\hat{\vec{y}})$$

where $\hat{\vec{y}} = \mathrm{softmax}(x)$ is a probability vector (the output of softmax on the class scores $\vec{x}$) and the vector $\vec{y}$ is a 1-hot encoded class label.

However, it's tricky to compute the gradient of softmax, so instead we'll define a version of cross-entropy that produces the exact same output but works directly on the class scores $\vec{x}$.

We can write, $$\begin{align} \ell_{\mathrm{CE}}(\vec{y},\hat{\vec{y}}) &= - {\vectr{y}} \log(\hat{\vec{y}}) = - {\vectr{y}} \log\left(\mathrm{softmax}(\vec{x})\right) \\ &= - {\vectr{y}} \log\left(\frac{e^{\vec{x}}}{\sum_k e^{x_k}}\right) \\ &= - \log\left(\frac{e^{x_y}}{\sum_k e^{x_k}}\right) \\ &= - \left(\log\left(e^{x_y}\right) - \log\left(\sum_k e^{x_k}\right)\right)\\ &= - x_y + \log\left(\sum_k e^{x_k}\right) \end{align}$$

Where the scalar $y$ is the correct class label, so $x_y$ is the correct class score.

Note that this version of cross entropy is also what's provided by PyTorch's nn module.

TODO: Complete the implementation of the CrossEntropyLoss class in the hw2/layers.py module.

In [11]:
# Test CrossEntropy
cross_entropy = layers.CrossEntropyLoss()
scores = torch.randn(N, num_classes)
labels = torch.randint(low=0, high=num_classes, size=(N,), dtype=torch.long)

# Test forward pass
loss = cross_entropy(scores, labels)
expected_loss = torch.nn.functional.cross_entropy(scores, labels)
test.assertLess(torch.abs(expected_loss-loss).item(), 1e-5)
print('loss=', loss.item())

# Test backward pass
test_block_grad(cross_entropy, scores, y=labels)
loss= 2.7283618450164795
Comparing gradients... 
input    diff=0.000

Building Models¶

Now that we have some working Layers, we can build an MLP model of arbitrary depth and compute end-to-end gradients.

First, lets copy an idea from PyTorch and implement our own version of the nn.Sequential Module. This is a Layer which contains other Layers and calls them in sequence. We'll use this to build our MLP model.

TODO: Complete the implementation of the Sequential class in the hw2/layers.py module.

In [12]:
# Test Sequential
# Let's create a long sequence of layers and see
# whether we can compute end-to-end gradients of the whole thing.

seq = layers.Sequential(
    layers.Linear(in_features, 100),
    layers.Linear(100, 200),
    layers.Linear(200, 100),
    layers.ReLU(),
    layers.Linear(100, 500),
    layers.LeakyReLU(alpha=0.01),
    layers.Linear(500, 200),
    layers.ReLU(),
    layers.Linear(200, 500),
    layers.LeakyReLU(alpha=0.1),
    layers.Linear(500, 1),
    layers.Sigmoid()
)
x_test = torch.randn(N, in_features)

# Test forward pass
z = seq(x_test)
test.assertSequenceEqual(z.shape, [N, 1])

# Test backward pass
test_block_grad(seq, x_test)
Comparing gradients... 
input    diff=0.000
param#01 diff=0.000
param#02 diff=0.000
param#03 diff=0.000
param#04 diff=0.000
param#05 diff=0.000
param#06 diff=0.000
param#07 diff=0.000
param#08 diff=0.000
param#09 diff=0.000
param#10 diff=0.000
param#11 diff=0.000
param#12 diff=0.000
param#13 diff=0.000
param#14 diff=0.000

Now, equipped with a Sequential, all we have to do is create an MLP architecture. We'll define our MLP with the following hyperparameters:

  • Number of input features, $D$.
  • Number of output classes, $C$.
  • Sizes of hidden layers, $h_1,\dots,h_L$.

So the architecture will be:

FC($D$, $h_1$) $\rightarrow$ ReLU $\rightarrow$ FC($h_1$, $h_2$) $\rightarrow$ ReLU $\rightarrow$ $\cdots$ $\rightarrow$ FC($h_{L-1}$, $h_L$) $\rightarrow$ ReLU $\rightarrow$ FC($h_{L}$, $C$)

We'll also create a sequence of the above MLP and a cross-entropy loss, since it's the gradient of the loss that we need in order to train a model.

TODO: Complete the implementation of the MLP class in the hw2/layers.py module. Ignore the dropout parameter for now.

In [13]:
# Create an MLP model
mlp = layers.MLP(in_features, num_classes, hidden_features=[100, 50, 100])
print(mlp)
MLP, Sequential
	[0] Linear(200, 100)
	[1] ReLU
	[2] Linear(100, 50)
	[3] ReLU
	[4] Linear(50, 100)
	[5] ReLU
	[6] Linear(100, 10)

In [14]:
# Test MLP architecture
N = 100
in_features = 10
num_classes = 10
for activation in ('relu', 'sigmoid'):
    mlp = layers.MLP(in_features, num_classes, hidden_features=[100, 50, 100], activation=activation)
    test.assertEqual(len(mlp.sequence), 7)
    
    num_linear = 0
    for b1, b2 in zip(mlp.sequence, mlp.sequence[1:]):
        if (str(b2).lower() == activation):
            test.assertTrue(str(b1).startswith('Linear'))
            num_linear += 1
            
    test.assertTrue(str(mlp.sequence[-1]).startswith('Linear'))
    test.assertEqual(num_linear, 3)

    # Test MLP gradients
    # Test forward pass
    x_test = torch.randn(N, in_features)
    labels = torch.randint(low=0, high=num_classes, size=(N,), dtype=torch.long)
    z = mlp(x_test)
    test.assertSequenceEqual(z.shape, [N, num_classes])

    # Create a sequence of MLPs and CE loss
    seq_mlp = layers.Sequential(mlp, layers.CrossEntropyLoss())
    loss = seq_mlp(x_test, y=labels)
    test.assertEqual(loss.dim(), 0)
    print(f'MLP loss={loss}, activation={activation}')

    # Test backward pass
    test_block_grad(seq_mlp, x_test, y=labels)
MLP loss=2.30924391746521, activation=relu
Comparing gradients... 
input    diff=0.000
param#01 diff=0.000
param#02 diff=0.000
param#03 diff=0.000
param#04 diff=0.000
param#05 diff=0.000
param#06 diff=0.000
param#07 diff=0.000
param#08 diff=0.000
MLP loss=2.3934402465820312, activation=sigmoid
Comparing gradients... 
input    diff=0.000
param#01 diff=0.000
param#02 diff=0.000
param#03 diff=0.000
param#04 diff=0.000
param#05 diff=0.000
param#06 diff=0.000
param#07 diff=0.000
param#08 diff=0.000

If the above tests passed then congratulations - you've now implemented an arbitrarily deep model and loss function with end-to-end automatic differentiation!

Questions¶

TODO Answer the following questions. Write your answers in the appropriate variables in the module hw2/answers.py.

In [15]:
from cs236781.answers import display_answer
import hw2.answers

Question 1¶

Suppose we have a linear (i.e. fully-connected) layer with a weight tensor $\mat{W}$, defined with in_features=1024 and out_features=512. We apply this layer to an input tensor $\mat{X}$ containing a batch of N=64 samples. The output of the layer is denoted as $\mat{Y}$.

  1. Consider the Jacobian tensor $\pderiv{\mat{Y}}{\mat{X}}$ of the output of the layer w.r.t. the input $\mat{X}$.

    1. What is the shape of this tensor?
    2. Is this Jacobian sparse (most elements zero by definition)? If so, why and which elements?
    3. Given the gradient of the output w.r.t. some downstream scalar loss $L$, $\delta\mat{Y}=\pderiv{L}{\mat{Y}}$, do we need to materialize the above Jacobian in order to calculate the downstream gratdient w.r.t. to the input ($\delta\mat{X}$)? If yes, explain why; if no, show how to calcualte it without materializing the Jacobian.
  2. Consider the Jacobian tensor $\pderiv{\mat{Y}}{\mat{W}}$ of the output of the layer w.r.t. the layer weights $\mat{W}$. Answer questions A-C about it as well.

In [16]:
display_answer(hw2.answers.part1_q1)

1. a) The shape of the gradient is (batch_size, out_features, batch_size, in_features) so in our case it's (64, 512, 64, 1024)

b) The above gradient is a tensor of shape (64, 512) where each item in it is a (64,1024) tensor which is the gradient of $X^TW$ according to the item (i,j) of X i.e $\frac{\partial Y}{\partial x_{i,j}} = \frac{\partial X^TW}{\partial x_{i,j}}$. But only a few items of the product depend on X(i,j) for each i,j. So in other words, most of the item of the gradient are 0.

To be more precise in $ Y_{m,n}$ only the only j-th row of $Y$ depends on $x_{i,j}$

Consquently in $(\frac{\partial Y}{\partial x_{i,j}})_{m,n}$ only the the j-th row is non-zero which means that every other entry is zero

c)No we don't need to compute the whole gradient, the trick is to use (his majesty) the chain rule. If each function is able to calculate its own derivative we just have to multiply them along the backward path

As we know $\delta\mat{X}=\pderiv{L}{\mat{X}} = \pderiv{L}{\mat{Y}}\pderiv{Y}{\mat{X}}$ But remember $Y$ is a Linear Layer which means $Y = X^T W + B$ So the gradient of $Y$ w.r.t to $X$ is easy to compute because it's only $W$

So we have $\delta\mat{X}=\pderiv{L}{\mat{Y}}W$ Where $\pderiv{L}{\mat{Y}}$ and $W$ are known

This method allows use to gain in time and space complexity.

-Time because with chain rule we don't have to compute a (64, 512, 64, 1024) tensor

-space because we won't have to store a (64, 512, 64, 1024) tensor in memory

2. a) The dimension of the Jacobian of the loss according to $W$ is (64,512,512,1024)

b) The above gradient is a tensor of shape (64, 512) where each item in it is a (512,1024) tensor which is the gradient of $X^TW$ according to the item (i,j) of W i.e $\frac{\partial Y}{\partial W_{i,j}} = \frac{\partial X^TW}{\partial W_{i,j}}$. But only a few items of the product depend on X(i,j) for each i,j. So in other words, most of the item of the gradient are 0.

To be more precise in $Y_{m,n}$ only the only i-th column of $Y$ depends on $W_{i,j}$

Consquently in $(\frac{\partial Y}{\partial W_{i,j}})_{m,n}$ only the the i-th column is non-zero which means that every other entry is zero.

c) No we don't need to compute the whole gradient, the trick is to use (his majesty) the chain rule. If each function is able to calculate its own derivative we just have to multiply them along the backward path

As we know $\delta\mat{W}=\pderiv{L}{\mat{X}} = \pderiv{L}{\mat{Y}}\pderiv{Y}{\mat{X}}$ But remember $Y$ is a Linear Layer which means $Y=X^TW+B$ So the gradient of $Y$ w.r.t to $W$ is easy to compute because it's only $X^^$

So we have $\delta W= (\frac{∂L}{∂Y})^T X$ Where $\frac{∂L}{∂Y}$ and $X$ are known

This method allows use to gain in time and space complexity.

-Time because with chain rule we don't have to compute a (64, 512, 512, 1024) tensor

-space because we won't have to store a (64, 512, 512, 1024) tensor in memory

Question 2¶

Is back-propagation required in order to train neural networks with decent-based optimization? Why or why not?

In [17]:
display_answer(hw2.answers.part1_q2)

Decent based algorithms need the gradient in order to make a step along the slope. Back-propagation is really popular in deep learning because it's an efficient (in terms of time and space) way to compute the gradient of the output according to one parameter, it doesn't require to compute the whole Jacobian which can have billions of entry and consequently make optimization more efficient. However as we have seen above computing the gradient of the loss directly according to a parameter is possible but much harder. But to answer to the question yes it's possible to use decent-based algorithms for optimization without using back-propagation. However by using backpropagation, we can make use of automatic differentiation which make the design of NN easier and faster

$$ \newcommand{\mat}[1]{\boldsymbol {#1}} \newcommand{\mattr}[1]{\boldsymbol {#1}^\top} \newcommand{\matinv}[1]{\boldsymbol {#1}^{-1}} \newcommand{\vec}[1]{\boldsymbol {#1}} \newcommand{\vectr}[1]{\boldsymbol {#1}^\top} \newcommand{\rvar}[1]{\mathrm {#1}} \newcommand{\rvec}[1]{\boldsymbol{\mathrm{#1}}} \newcommand{\diag}{\mathop{\mathrm {diag}}} \newcommand{\set}[1]{\mathbb {#1}} \newcommand{\norm}[1]{\left\lVert#1\right\rVert} \newcommand{\pderiv}[2]{\frac{\partial #1}{\partial #2}} \newcommand{\bb}[1]{\boldsymbol{#1}} $$

Part 2: Optimization and Training¶

In this part we will learn how to implement optimization algorithms for deep networks. Additionally, we'll learn how to write training loops and implement a modular model trainer. We'll use our optimizers and training code to test a few configurations for classifying images with an MLP model.

In [1]:
import os
import numpy as np
import matplotlib.pyplot as plt
import unittest
import torch
import torchvision
import torchvision.transforms as tvtf

%matplotlib inline
%load_ext autoreload
%autoreload 2
In [2]:
seed = 42
plt.rcParams.update({'font.size': 12})
test = unittest.TestCase()

Implementing Optimization Algorithms¶

In the context of deep learning, an optimization algorithm is some method of iteratively updating model parameters so that the loss converges toward some local minimum (which we hope will be good enough).

Gradient descent-based methods are by far the most popular algorithms for optimization of neural network parameters. However the high-dimensional loss-surfaces we encounter in deep learning applications are highly non-convex. They may be riddled with local minima, saddle points, large plateaus and a host of very challenging "terrain" for gradient-based optimization. This gave rise to many different methods of performing the parameter updates based on the loss gradients, aiming to tackle these optimization challenges.

The most basic gradient-based update rule can be written as,

$$ \vec{\theta} \leftarrow \vec{\theta} - \eta \nabla_{\vec{\theta}} L(\vec{\theta}; \mathcal{D}) $$

where $\mathcal{D} = \left\{ (\vec{x}^i, \vec{y}^i) \right\}_{i=1}^{M}$ is our training dataset or part of it. Specifically, if we have in total $N$ training samples, then

  • If $M=N$ this is known as regular gradient descent. If the dataset does not fit in memory the gradient of this loss becomes infeasible to compute.
  • If $M=1$, the loss is computed w.r.t. a single different sample each time. This is known as stochastic gradient descent.
  • If $1<M<N$ this is known as stochastic mini-batch gradient descent. This is the most commonly-used option.

The intuition behind gradient descent is simple: since the gradient of a multivariate function points to the direction of steepest ascent ("uphill"), we move in the opposite direction. A small step size $\eta$ known as the learning rate is required since the gradient can only serve as a first-order linear approximation of the function's behaviour at $\vec{\theta}$ (recall e.g. the Taylor expansion). However in truth our loss surface generally has nontrivial curvature caused by a high order nonlinear dependency on $\vec{\theta}$. Thus taking a large step in the direction of the gradient is actually just as likely to increase the function value.

The idea behind the stochastic versions is that by constantly changing the samples we compute the loss with, we get a dynamic error surface, i.e. it's different for each set of training samples. This is thought to generally improve the optimization since it may help the optimizer get out of flat regions or sharp local minima since these features may disappear in the loss surface of subsequent batches. The image below illustrates this. The different lines are different 1-dimensional losses for different training set-samples.

Deep learning frameworks generally provide implementations of various gradient-based optimization algorithms. Here we'll implement our own optimization module from scratch, this time keeping a similar API to the PyTorch optim package.

We define a base Optimizer class. An optimizer holds a set of parameter tensors (these are the trainable parameters of some model) and maintains internal state. It may be used as follows:

  • After the forward pass has been performed the optimizer's zero_grad() function is invoked to clear the parameter gradients computed by previous iterations.
  • After the backward pass has been performed, and gradients have been calculated for these parameters, the optimizer's step() function is invoked in order to update the value of each parameter based on it's gradient.

The exact method of update is implementation-specific for each optimizer and may depend on its internal state. In addition, adding the regularization penalty to the gradient is handled by the optimizer since it only depends on the parameter values (and not the data).

Here's the API of our Optimizer:

In [3]:
import hw2.optimizers as optimizers
help(optimizers.Optimizer)
Help on class Optimizer in module hw2.optimizers:

class Optimizer(abc.ABC)
 |  Optimizer(params)
 |  
 |  Base class for optimizers.
 |  
 |  Method resolution order:
 |      Optimizer
 |      abc.ABC
 |      builtins.object
 |  
 |  Methods defined here:
 |  
 |  __init__(self, params)
 |      :param params: A sequence of model parameters to optimize. Can be a
 |      list of (param,grad) tuples as returned by the Layers, or a list of
 |      pytorch tensors in which case the grad will be taken from them.
 |  
 |  step(self)
 |      Updates all the registered parameter values based on their gradients.
 |  
 |  zero_grad(self)
 |      Sets the gradient of the optimized parameters to zero (in place).
 |  
 |  ----------------------------------------------------------------------
 |  Data descriptors defined here:
 |  
 |  __dict__
 |      dictionary for instance variables (if defined)
 |  
 |  __weakref__
 |      list of weak references to the object (if defined)
 |  
 |  params
 |      :return: A sequence of parameter tuples, each tuple containing
 |      (param_data, param_grad). The data should be updated in-place
 |      according to the grad.
 |  
 |  ----------------------------------------------------------------------
 |  Data and other attributes defined here:
 |  
 |  __abstractmethods__ = frozenset({'step'})

Vanilla SGD with Regularization¶

Let's start by implementing the simplest gradient based optimizer. The update rule will be exacly as stated above, but we'll also add a L2-regularization term to the gradient. Remember that in the loss function, the L2 regularization term is expressed by

$$R(\vec{\theta}) = \frac{1}{2}\lambda||\vec{\theta}||^2_2.$$

TODO: Complete the implementation of the VanillaSGD class in the hw2/optimizers.py module.

In [4]:
# Test VanillaSGD
torch.manual_seed(42)
p = torch.randn(500, 10)
dp = torch.randn(*p.shape)*2
params = [(p, dp)]

vsgd = optimizers.VanillaSGD(params, learn_rate=0.5, reg=0.1)
vsgd.step()

expected_p = torch.load('tests/assets/expected_vsgd.pt')
diff = torch.norm(p-expected_p).item()
print(f'diff={diff}')
test.assertLess(diff, 1e-3)
diff=1.7634911273489706e-05

Training¶

Now that we can build a model and loss function, compute their gradients and we have an optimizer, we can finally do some training!

In the spirit of more modular software design, we'll implement a class that will aid us in automating the repetitive training loop code that we usually write over and over again. This will be useful for both training our Layer-based models and also later for training PyTorch nn.Modules.

Here's our Trainer API:

In [5]:
import hw2.training as training
help(training.Trainer)
Help on class Trainer in module hw2.training:

class Trainer(abc.ABC)
 |  Trainer(model: torch.nn.modules.module.Module, device: Union[torch.device, NoneType] = None)
 |  
 |  A class abstracting the various tasks of training models.
 |  
 |  Provides methods at multiple levels of granularity:
 |  - Multiple epochs (fit)
 |  - Single epoch (train_epoch/test_epoch)
 |  - Single batch (train_batch/test_batch)
 |  
 |  Method resolution order:
 |      Trainer
 |      abc.ABC
 |      builtins.object
 |  
 |  Methods defined here:
 |  
 |  __init__(self, model: torch.nn.modules.module.Module, device: Union[torch.device, NoneType] = None)
 |      Initialize the trainer.
 |      :param model: Instance of the model to train.
 |      :param device: torch.device to run training on (CPU or GPU).
 |  
 |  fit(self, dl_train: torch.utils.data.dataloader.DataLoader, dl_test: torch.utils.data.dataloader.DataLoader, num_epochs: int, checkpoints: str = None, early_stopping: int = None, print_every: int = 1, **kw) -> cs236781.train_results.FitResult
 |      Trains the model for multiple epochs with a given training set,
 |      and calculates validation loss over a given validation set.
 |      :param dl_train: Dataloader for the training set.
 |      :param dl_test: Dataloader for the test set.
 |      :param num_epochs: Number of epochs to train for.
 |      :param checkpoints: Whether to save model to file every time the
 |          test set accuracy improves. Should be a string containing a
 |          filename without extension.
 |      :param early_stopping: Whether to stop training early if there is no
 |          test loss improvement for this number of epochs.
 |      :param print_every: Print progress every this number of epochs.
 |      :return: A FitResult object containing train and test losses per epoch.
 |  
 |  save_checkpoint(self, checkpoint_filename: str)
 |      Saves the model in it's current state to a file with the given name (treated
 |      as a relative path).
 |      :param checkpoint_filename: File name or relative path to save to.
 |  
 |  test_batch(self, batch) -> cs236781.train_results.BatchResult
 |      Runs a single batch forward through the model and calculates loss.
 |      :param batch: A single batch of data  from a data loader (might
 |          be a tuple of data and labels or anything else depending on
 |          the underlying dataset.
 |      :return: A BatchResult containing the value of the loss function and
 |          the number of correctly classified samples in the batch.
 |  
 |  test_epoch(self, dl_test: torch.utils.data.dataloader.DataLoader, **kw) -> cs236781.train_results.EpochResult
 |      Evaluate model once over a test set (single epoch).
 |      :param dl_test: DataLoader for the test set.
 |      :param kw: Keyword args supported by _foreach_batch.
 |      :return: An EpochResult for the epoch.
 |  
 |  train_batch(self, batch) -> cs236781.train_results.BatchResult
 |      Runs a single batch forward through the model, calculates loss,
 |      preforms back-propagation and updates weights.
 |      :param batch: A single batch of data  from a data loader (might
 |          be a tuple of data and labels or anything else depending on
 |          the underlying dataset.
 |      :return: A BatchResult containing the value of the loss function and
 |          the number of correctly classified samples in the batch.
 |  
 |  train_epoch(self, dl_train: torch.utils.data.dataloader.DataLoader, **kw) -> cs236781.train_results.EpochResult
 |      Train once over a training set (single epoch).
 |      :param dl_train: DataLoader for the training set.
 |      :param kw: Keyword args supported by _foreach_batch.
 |      :return: An EpochResult for the epoch.
 |  
 |  ----------------------------------------------------------------------
 |  Data descriptors defined here:
 |  
 |  __dict__
 |      dictionary for instance variables (if defined)
 |  
 |  __weakref__
 |      list of weak references to the object (if defined)
 |  
 |  ----------------------------------------------------------------------
 |  Data and other attributes defined here:
 |  
 |  __abstractmethods__ = frozenset({'test_batch', 'train_batch'})

The Trainer class splits the task of training (and evaluating) models into three conceptual levels,

  • Multiple epochs - the fit method, which returns a FitResult containing losses and accuracies for all epochs.
  • Single epoch - the train_epoch and test_epoch methods, which return an EpochResult containing losses per batch and the single accuracy result of the epoch.
  • Single batch - the train_batch and test_batch methods, which return a BatchResult containing a single loss and the number of correctly classified samples in the batch.

It implements the first two levels. Inheriting classes are expected to implement the single-batch level methods since these are model and/or task specific.

The first thing we should do in order to verify our model, gradient calculations and optimizer implementation is to try to overfit a large model (many parameters) to a small dataset (few images). This will show us that things are working properly.

Let's begin by loading the CIFAR-10 dataset.

In [6]:
data_dir = os.path.expanduser('~/.pytorch-datasets')
ds_train = torchvision.datasets.CIFAR10(root=data_dir, download=True, train=True, transform=tvtf.ToTensor())
ds_test = torchvision.datasets.CIFAR10(root=data_dir, download=True, train=False, transform=tvtf.ToTensor())

print(f'Train: {len(ds_train)} samples')
print(f'Test: {len(ds_test)} samples')
Files already downloaded and verified
Files already downloaded and verified
Train: 50000 samples
Test: 10000 samples

Now, let's implement just a small part of our training logic since that's what we need right now.

TODO:

  1. Complete the implementation of the train_batch() method in the LayerTrainer class within the hw2/training.py module.
  2. Update the hyperparameter values in the part2_overfit_hp() function in the hw2/answers.py module. Tweak the hyperparameter values until your model overfits a small number of samples in the code block below. You should get 100% accuracy within a few epochs.

The following code block will use your custom Layer-based MLP implentation, custom Vanilla SGD and custom trainer to overfit the data. The classification accuracy should be 100% within a few epochs.

In [7]:
import hw2.layers as layers
import hw2.answers as answers
from torch.utils.data import DataLoader

# Overfit to a very small dataset of 20 samples
batch_size = 10
max_batches = 2
dl_train = torch.utils.data.DataLoader(ds_train, batch_size, shuffle=False)

# Get hyperparameters
hp = answers.part2_overfit_hp()

torch.manual_seed(seed)

# Build a model and loss using our custom MLP and CE implementations
model = layers.MLP(3*32*32, num_classes=10, hidden_features=[128]*3, wstd=hp['wstd'])
loss_fn = layers.CrossEntropyLoss()

# Use our custom optimizer
optimizer = optimizers.VanillaSGD(model.params(), learn_rate=hp['lr'], reg=hp['reg'])

# Run training over small dataset multiple times
trainer = training.LayerTrainer(model, loss_fn, optimizer)
best_acc = 0
for i in range(20):
    res = trainer.train_epoch(dl_train, max_batches=max_batches)
    best_acc = res.accuracy if res.accuracy > best_acc else best_acc
    
test.assertGreaterEqual(best_acc, 98)
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]
train_batch:   0%|          | 0/2 [00:00<?, ?it/s]

Now that we know training works, let's try to fit a model to a bit more data for a few epochs, to see how well we're doing. First, we need a function to plot the FitResults object.

In [8]:
from cs236781.plot import plot_fit
plot_fit?

TODO:

  1. Complete the implementation of the test_batch() method in the LayerTrainer class within the hw2/training.py module.
  2. Implement the fit() method of the Trainer class within the hw2/training.py module.
  3. Tweak the hyperparameters for this section in the part2_optim_hp() function in the hw2/answers.py module.
  4. Run the following code blocks to train. Try to get above 35-40% test-set accuracy.
In [9]:
# Define a larger part of the CIFAR-10 dataset (still not the whole thing)
batch_size = 50
max_batches = 100
in_features = 3*32*32
num_classes = 10
dl_train = torch.utils.data.DataLoader(ds_train, batch_size, shuffle=False)
dl_test = torch.utils.data.DataLoader(ds_test, batch_size//2, shuffle=False)
In [10]:
# Define a function to train a model with our Trainer and various optimizers
def train_with_optimizer(opt_name, opt_class, fig):
    torch.manual_seed(seed)
    
    # Get hyperparameters
    hp = answers.part2_optim_hp()
    hidden_features = [128] * 5
    num_epochs = 10
    
    # Create model, loss and optimizer instances
    model = layers.MLP(in_features, num_classes, hidden_features, wstd=hp['wstd'])
    loss_fn = layers.CrossEntropyLoss()
    optimizer = opt_class(model.params(), learn_rate=hp[f'lr_{opt_name}'], reg=hp['reg'])

    # Train with the Trainer
    trainer = training.LayerTrainer(model, loss_fn, optimizer)
    fit_res = trainer.fit(dl_train, dl_test, num_epochs, max_batches=max_batches)
    
    fig, axes = plot_fit(fit_res, fig=fig, legend=opt_name)
    return fig
In [11]:
fig_optim = None
fig_optim = train_with_optimizer('vanilla', optimizers.VanillaSGD, fig_optim)
--- EPOCH 1/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 2/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 3/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 4/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 5/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 6/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 7/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 8/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 9/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 10/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]

Momentum¶

The simple vanilla SGD update is rarely used in practice since it's very slow to converge relative to other optimization algorithms.

One reason is that naïvely updating in the direction of the current gradient causes it to fluctuate wildly in areas where the loss surface in some dimensions is much steeper than in others. Another reason is that using the same learning rate for all parameters is not a great idea since not all parameters are created equal. For example, parameters associated with rare features should be updated with a larger step than ones associated with commonly-occurring features because they'll get less updates through the gradients.

Therefore more advanced optimizers take into account the previous gradients of a parameter and/or try to use a per-parameter specific learning rate instead of a common one.

Let's now implement a simple and common optimizer: SGD with Momentum. This optimizer takes previous gradients of a parameter into account when updating it's value instead of just the current one. In practice it usually provides faster convergence than the vanilla SGD.

The SGD with Momentum update rule can be stated as follows: $$\begin{align} \vec{v}_{t+1} &= \mu \vec{v}_t - \eta \delta \vec{\theta}_t \\ \vec{\theta}_{t+1} &= \vec{\theta}_t + \vec{v}_{t+1} \end{align}$$

Where $\eta$ is the learning rate, $\vec{\theta}$ is a model parameter, $\delta \vec{\theta}_t=\pderiv{L}{\vec{\theta}}(\vec{\theta}_t)$ is the gradient of the loss w.r.t. to the parameter and $0\leq\mu<1$ is a hyperparameter known as momentum.

Expanding the update rule recursively shows us now the parameter update infact depends on all previous gradient values for that parameter, where the old gradients are exponentially decayed by a factor of $\mu$ at each timestep.

Since we're incorporating previous gradient (update directions), a noisy value of the current gradient will have less effect so that the general direction of previous updates is maintained somewhat. The following figure illustrates this.

TODO:

  1. Complete the implementation of the MomentumSGD class in the hw2/optimizers.py module.
  2. Tweak the learning rate for momentum in part2_optim_hp() the function in the hw2/answers.py module.
  3. Run the following code block to compare to the vanilla SGD.
In [12]:
fig_optim = train_with_optimizer('momentum', optimizers.MomentumSGD, fig_optim)
fig_optim
--- EPOCH 1/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 2/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 3/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 4/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 5/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 6/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 7/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 8/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 9/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 10/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
Out[12]:

RMSProp¶

This is another optmizer that accounts for previous gradients, but this time it uses them to adapt the learning rate per parameter.

RMSProp maintains a decaying moving average of previous squared gradients, $$ \vec{r}_{t+1} = \gamma\vec{r}_{t} + (1-\gamma)\delta\vec{\theta}_t^2 $$ where $0<\gamma<1$ is a decay constant usually set close to $1$, and $\delta\vec{\theta}_t^2$ denotes element-wise squaring.

The update rule for each parameter is then, $$ \vec{\theta}_{t+1} = \vec{\theta}_t - \left( \frac{\eta}{\sqrt{r_{t+1}+\varepsilon}} \right) \delta\vec{\theta}_t $$

where $\varepsilon$ is a small constant to prevent numerical instability. The idea here is to decrease the learning rate for parameters with high gradient values and vice-versa. The decaying moving average prevents accumulating all the past gradients which would cause the effective learning rate to become zero.

TODO:

  1. Complete the implementation of the RMSProp class in the hw2/optimizers.py module.
  2. Tweak the learning rate for RMSProp in part2_optim_hp() the function in the hw2/answers.py module.
  3. Run the following code block to compare to the other optimizers.
In [13]:
fig_optim = train_with_optimizer('rmsprop', optimizers.RMSProp, fig_optim)
fig_optim
--- EPOCH 1/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 2/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 3/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 4/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 5/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 6/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 7/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 8/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 9/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
--- EPOCH 10/10 ---
train_batch:   0%|          | 0/100 [00:00<?, ?it/s]
test_batch:   0%|          | 0/100 [00:00<?, ?it/s]
Out[13]:

Note that you should get better train/test accuracy with Momentum and RMSProp than Vanilla.

Dropout Regularization¶

Dropout is a useful technique to improve generalization of deep models.

The idea is simple: during the forward pass drop, i.e. set to to zero, the activation of each neuron, with a probability of $p$. For example, if $p=0.4$ this means we drop the activations of 40% of the neurons (on average).

There are a few important things to note about dropout:

  1. It is only performed during training. When testing our model the dropout layers should be a no-op.
  2. In the backward pass, gradients are only propagated back into neurons that weren't dropped during the forward pass.
  3. During testing, the activations must be scaled since the expected value of each neuron during the training phase is now $1-p$ times it's original expectation. Thus, we need to scale the test-time activations by $1-p$ to match. Equivalently, we can scale the train time activations by $1/(1-p)$.

TODO:

  1. Complete the implementation of the Dropout class in the hw2/layers.py module.
  2. Finish the implementation of the MLP's __init__() method in the hw2/layers.py module. If dropout>0 you should add a Dropout layer after each ReLU.
In [14]:
from hw2.grad_compare import compare_layer_to_torch

# Check architecture of MLP with dropout layers
mlp_dropout = layers.MLP(in_features, num_classes, [50]*3, dropout=0.6)
print(mlp_dropout)
test.assertEqual(len(mlp_dropout.sequence), 10)
for b1, b2 in zip(mlp_dropout.sequence, mlp_dropout.sequence[1:]):
    if str(b1).lower() == 'relu':
        test.assertTrue(str(b2).startswith('Dropout'))
test.assertTrue(str(mlp_dropout.sequence[-1]).startswith('Linear'))
MLP, Sequential
	[0] Linear(3072, 50)
	[1] ReLU
	[2] Dropout(p=0.6)
	[3] Linear(50, 50)
	[4] ReLU
	[5] Dropout(p=0.6)
	[6] Linear(50, 50)
	[7] ReLU
	[8] Dropout(p=0.6)
	[9] Linear(50, 10)

In [15]:
# Test end-to-end gradient in train and test modes.
print('Dropout, train mode')
mlp_dropout.train(True)
for diff in compare_layer_to_torch(mlp_dropout, torch.randn(500, in_features)):
    test.assertLess(diff, 1e-3)
    
print('Dropout, test mode')
mlp_dropout.train(False)
for diff in compare_layer_to_torch(mlp_dropout, torch.randn(500, in_features)):
    test.assertLess(diff, 1e-3)
Dropout, train mode
Comparing gradients... 
input    diff=0.000
param#01 diff=0.000
param#02 diff=0.000
param#03 diff=0.000
param#04 diff=0.000
param#05 diff=0.000
param#06 diff=0.000
param#07 diff=0.000
param#08 diff=0.000
Dropout, test mode
Comparing gradients... 
input    diff=0.000
param#01 diff=0.000
param#02 diff=0.000
param#03 diff=0.000
param#04 diff=0.000
param#05 diff=0.000
param#06 diff=0.000
param#07 diff=0.000
param#08 diff=0.000

To see whether dropout really improves generalization, let's take a small training set (small enough to overfit) and a large test set and check whether we get less overfitting and perhaps improved test-set accuracy when using dropout.

In [16]:
# Define a small set from CIFAR-10, but take a larger test set since we want to test generalization
batch_size = 10
max_batches = 40
in_features = 3*32*32
num_classes = 10
dl_train = torch.utils.data.DataLoader(ds_train, batch_size, shuffle=False)
dl_test = torch.utils.data.DataLoader(ds_test, batch_size*2, shuffle=False)

TODO: Tweak the hyperparameters for this section in the part2_dropout_hp() function in the hw2/answers.py module. Try to set them so that the first model (with dropout=0) overfits. You can disable the other dropout options until you tune the hyperparameters. We can then see the effect of dropout for generalization.

In [17]:
# Get hyperparameters
hp = answers.part2_dropout_hp()
hidden_features = [400] * 1
num_epochs = 30
In [18]:
torch.manual_seed(seed)
fig=None
#for dropout in [0]:  # Use this for tuning the hyperparms until you overfit
for dropout in [0, 0.4, 0.8]:
    model = layers.MLP(in_features, num_classes, hidden_features, wstd=hp['wstd'], dropout=dropout)
    loss_fn = layers.CrossEntropyLoss()
    optimizer = optimizers.MomentumSGD(model.params(), learn_rate=hp['lr'], reg=0)

    print('*** Training with dropout=', dropout)
    trainer = training.LayerTrainer(model, loss_fn, optimizer)
    fit_res_dropout = trainer.fit(dl_train, dl_test, num_epochs, max_batches=max_batches, print_every=6)
    fig, axes = plot_fit(fit_res_dropout, fig=fig, legend=f'dropout={dropout}', log_loss=True)
*** Training with dropout= 0
--- EPOCH 1/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 7/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 13/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 19/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 25/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 30/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
*** Training with dropout= 0.4
--- EPOCH 1/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 7/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 13/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 19/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 25/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 30/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
*** Training with dropout= 0.8
--- EPOCH 1/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 7/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 13/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 19/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 25/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]
--- EPOCH 30/30 ---
train_batch:   0%|          | 0/40 [00:00<?, ?it/s]
test_batch:   0%|          | 0/40 [00:00<?, ?it/s]

Questions¶

TODO Answer the following questions. Write your answers in the appropriate variables in the module hw2/answers.py.

In [19]:
from cs236781.answers import display_answer
import hw2.answers

Question 1¶

Regarding the graphs you got for the three dropout configurations:

  1. Explain the graphs of no-dropout vs dropout. Do they match what you expected to see?

    • If yes, explain why and provide examples based on the graphs.
    • If no, explain what you think the problem is and what should be modified to fix it.
  2. Compare the low-dropout setting to the high-dropout setting and explain based on your graphs.

In [20]:
display_answer(hw2.answers.part2_q1)

1) Yes the graphs match with our expectation. For the training accuracy, lower the dropout is higher the accuracy is, this make sens since less we miss information on our training set more our model will fit to the known data and then will be more accurate on the training. While the testing set contains only unknown data, so missing some of information about the training set will reduce the error due to overfitting

2)Having an high dropout setting will make more of our weight to 0, which means that we erase most of our knowledge about the data. So a too high dropout will reduce our performances. But a more reasonable dropout will make our classifier's decisions less based on the information based during training and consequently reduces the overfitting. As we can see on the graphs the training accuracy is higher when the dropout is low. However the best testing accuracy is obtained when the dropout is set to 0.4 which mean that indeed having some dropout reduces overfitting. Moreover when the dropout is to 0.8 both training and testing accuracy are low which is the pattern of underfitting. In fact with such a dropout we cancel 80% of our weights and then the model doesn't have enough knowledge to be accurate.

Question 2¶

When training a model with the cross-entropy loss function, is it possible for the test loss to increase for a few epochs while the test accuracy also increases?

If it's possible explain how, if it's not explain why not.

In [21]:
display_answer(hw2.answers.part2_q2)

Even if it seems that accuracy and loss are inversely correlated in the case of Cross Entropy Loss it is possible for the test loss to increase when the test accuracy also increases. Loss measures a difference between raw prediction and class while accuracy measures a difference between threshold prediction and class. Therefore if raw predictions changes, it directly influences the loss whereas in order to affect the accuracy the predictions has to go over or under a threshold. Also the loss function is continuous while accuracy is in a finite set of values. Consequently during some epochs ,where the model still predict well, for the loss to increase. So if a few raw predictions were bad but the model still predicts right it causes the loss to increase, while accuracy stays the same. At the same time the model is still learning patterns which are useful for generalization therefore increases accuracy.

Question 3¶

  1. Explain the difference between gradient descent and back-propagation.

  2. Compare in detail between gradient descent (GD) and stochastic gradient descent (SGD).

  3. Why is SGD used more often in the practice of deep learning? Provide a few justifications.

  4. You would like to try GD to train your model instead of SGD, but you're concerned that your dataset won't fit in memory. A friend suggested that you should split the data into disjoint batches, do multiple forward passes until all data is exhausted, and then do one backward pass on the sum of the losses.

    1. Would this approach produce a gradient equivalent to GD? Why or why not? provide mathematical justification for your answer.
    2. You implemented the suggested approach, and were careful to use batch sizes small enough so that each batch fits in memory. However, after some number of batches you got an out of memory error. What happened?
In [22]:
display_answer(hw2.answers.part2_q3)

1 Back propagation compute the gradient w.r.t to every parameter using chain rule along the backward path while gradient descent is the algorithm that optimize our parameters by computing the gradient in order to get to the minimum of the function

2 During gradient descent we compute the exact gradient to reduce the loss at each iteration. In order to do that we compute the gradient over all data points and then we average them:

$\nabla _wL = \frac{1}{m}∑^m_i \frac{∂L(x^{(i)})}{\partial w}$

Also the loss converges with a linear rate with GD.

While in Stochastic Gradient Descent we average the gradient w.r.t to the parameter by computing the gradient on one small set of points $C \subseteq S$ where the samples and the size of $C$ are randoms.

In SGD:

$\nabla _wL \approx \frac{1}{|C|}∑_{x ⊆C} \frac{∂L(x)}{\partial w}$

Then in both algorithm update the weights like below

$w ← w - \eta \nabla_w L$ Where $\eta \in ℝ$ is the learning rate

3 In practice SGD is more used in Deep Learning for the following two reasons:

1) Computing the exact gradient of the loss according to all the training points is really long knowing that in neural network architectures we often have millions of parameters and in training set we have also thousands of samples so for instance if we have $10^6$ parameters and $10^4$ training samples we get $10^{10}$ partial derivatives to compute at each iterations.

2) GD also suffers from the local minimum problem, indeed if the direction of the gradient leads to a local minimum GD will get "trapped" in it whereas computing a random gradient can avoid this kind of problems

4

1) This method isn't exactly as Gradient Descent, in GD the gradient is computed by summing the loss according to all point in the dataset while here the gradient is computed on the sum of the loss. In fact knowing that the gradient is linear $(\nabla \sum f = \sum \nabla f)$ we get that the gradient of GD multiplied by the number of samples. So we won't have the scaling factor $frac{1}{m}$. But the gradients are equivalent

2)In the back-propagation method all of the layers need to compute the derivative of their output w.r.t. to their input. Conequently all of the layers of the MLP need to store their input. The out of memory error happened because of the accumulation of all of the inputs in all of the layers. Hence, while the forward layer was computed by loading the data in batches, the layers stockpiled all of the data-points that they have encountered which lead to the out of memory error.

In [22]:
 
$$ \newcommand{\mat}[1]{\boldsymbol {#1}} \newcommand{\mattr}[1]{\boldsymbol {#1}^\top} \newcommand{\matinv}[1]{\boldsymbol {#1}^{-1}} \newcommand{\vec}[1]{\boldsymbol {#1}} \newcommand{\vectr}[1]{\boldsymbol {#1}^\top} \newcommand{\rvar}[1]{\mathrm {#1}} \newcommand{\rvec}[1]{\boldsymbol{\mathrm{#1}}} \newcommand{\diag}{\mathop{\mathrm {diag}}} \newcommand{\set}[1]{\mathbb {#1}} \newcommand{\norm}[1]{\left\lVert#1\right\rVert} \newcommand{\pderiv}[2]{\frac{\partial #1}{\partial #2}} \newcommand{\bb}[1]{\boldsymbol{#1}} $$

Part 3: Binary Classification with Multilayer Perceptrons¶

In this part we'll implement a general purpose MLP and Binary Classifier using pytorch. We'll implement its training, and also learn about decision boundaries an threshold selection in the context of binary classification. Finally, we'll explore the effect of depth and width on an MLP's performance.

In [32]:
import os
import re
import sys
import glob
import unittest
from typing import Sequence, Tuple

import sklearn
import numpy as np
import matplotlib.pyplot as plt
import torch
import torchvision
import torch.nn as nn
import torchvision.transforms as tvtf
from torch import Tensor

%matplotlib inline
%load_ext autoreload
%autoreload 2
The autoreload extension is already loaded. To reload it, use:
  %reload_ext autoreload
In [33]:
seed = 42
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
plt.rcParams.update({'font.size': 12})
test = unittest.TestCase()

Synthetic Dataset¶

To test our first neural network-based classifiers we'll start by creating a toy binary classification dataset, but one which is not trivial for a linear model.

In [34]:
from sklearn.datasets import make_moons
from sklearn.model_selection import train_test_split
In [35]:
def rotate_2d(X, deg=0):
    """
    Rotates each 2d sample in X of shape (N, 2) by deg degrees.
    """
    a = np.deg2rad(deg)
    return X @ np.array([[np.cos(a), -np.sin(a)],[np.sin(a), np.cos(a)]]).T

def plot_dataset_2d(X, y, n_classes=2, alpha=0.2, figsize=(8, 6), title=None, ax=None):
    if ax is None:
        fig, ax = plt.subplots(1, 1, figsize=figsize)
    for c in range(n_classes):
        ax.scatter(*X[y==c,:].T, alpha=alpha, label=f"class {c}");
        
    ax.set_xlabel("$x_1$"); ax.set_ylabel("$x_2$");
    ax.legend(); ax.set_title((title or '') + f" (n={len(y)})")

We'll split our data into 80% train and validation, and 20% test. To make it a bit more challenging, we'll simulate a somewhat real-world setting where there are multiple populations, and the training/validation data is not sampled iid from the underlying data distribution.

In [36]:
np.random.seed(seed)

N = 10_000
N_train = int(N * .8)

# Create data from two different distributions for the training/validation
X1, y1 = make_moons(n_samples=N_train//2, noise=0.2)
X1 = rotate_2d(X1, deg=10)
X2, y2 = make_moons(n_samples=N_train//2, noise=0.25)
X2 = rotate_2d(X2, deg=50)

# Test data comes from a similar but noisier distribution
X3, y3 = make_moons(n_samples=(N-N_train), noise=0.3)
X3 = rotate_2d(X3, deg=40)

X, y = np.vstack([X1, X2, X3]), np.hstack([y1, y2, y3])
In [37]:
# Train and validation data is from mixture distribution
X_train, X_valid, y_train, y_valid = train_test_split(X[:N_train, :], y[:N_train], test_size=1/3, shuffle=False)

# Test data is only from the second distribution
X_test, y_test = X[N_train:, :], y[N_train:]

fig, ax = plt.subplots(1, 3, figsize=(20, 5))
plot_dataset_2d(X_train, y_train, title='Train', ax=ax[0]);
plot_dataset_2d(X_valid, y_valid, title='Validation', ax=ax[1]);
plot_dataset_2d(X_test, y_test, title='Test', ax=ax[2]);

Now let us create a data loader for each dataset.

In [38]:
from torch.utils.data import TensorDataset
from torch.utils.data import DataLoader

batch_size = 32

dl_train, dl_valid, dl_test = [
    DataLoader(
        dataset=TensorDataset(
            torch.from_numpy(X_).to(torch.float32),
            torch.from_numpy(y_)
        ),
        shuffle=True,
        num_workers=0,
        batch_size=batch_size
    )
    for X_, y_ in [(X_train, y_train), (X_valid, y_valid), (X_test, y_test)]
]

print(f'{len(dl_train.dataset)}, {len(dl_valid.dataset)}, {len(dl_test.dataset)}')
5333, 2667, 2000

Simple MLP¶

A multilayer-perceptron is arguably a the most basic type of neural network model. It is composed of $L$ layers, each layer $l$ with $n_l$ perceptron ("neuron") units. Each perceptron is connected to all ouputs of the previous layer (or all inputs in the first layer), calculates their weighted sum, applies a linearity and produces a single output.

Each layer $l$ operates on the output of the previous layer ($\vec{y}_{l-1}$) and calculates:

$$ \vec{y}_l = \varphi\left( \mat{W}_l \vec{y}_{l-1} + \vec{b}_l \right),~ \mat{W}_l\in\set{R}^{n_{l}\times n_{l-1}},~ \vec{b}_l\in\set{R}^{n_l},~ l \in \{1,2,\dots,L\}. $$
  • Note that both input and output are vectors. We can think of the above equation as describing a layer of multiple perceptrons.
  • We'll henceforth refer to such layers as fully-connected or FC layers.
  • The first layer accepts the input of the model, i.e. $\vec{y}_0=\vec{x}\in\set{R}^d$.
  • The last layer, $L$, is the output layer, so $y_L$ is the output of the model.
  • The layers $1, 2, \dots, L-1$ are called hidden layers.

To begin, let's implement a general multi-layer perceptron model. We'll seek to implement it in a way which is both general in terms of architecture, and also composable so that we can use our MLP in the context of larger models.

TODO: Implement the MLP class in the hw2/mlp.py module.

In [39]:
from hw2.mlp import MLP

mlp = MLP(
    in_dim=2,
    dims=[8, 16, 32, 64],
    nonlins=['relu', 'tanh', nn.LeakyReLU(0.314), 'softmax']
)
mlp
Out[39]:
MLP(
  (seq): Sequential(
    (0): Linear(in_features=2, out_features=8, bias=True)
    (1): ReLU()
    (2): Linear(in_features=8, out_features=16, bias=True)
    (3): Tanh()
    (4): Linear(in_features=16, out_features=32, bias=True)
    (5): LeakyReLU(negative_slope=0.314)
    (6): Linear(in_features=32, out_features=64, bias=True)
    (7): Softmax(dim=None)
  )
)

Let's try our implementation on a batch of data.

In [40]:
x0, y0 = next(iter(dl_train))

yhat0 = mlp(x0)

test.assertEqual(len([*mlp.parameters()]), 8)
test.assertEqual(yhat0.shape, (batch_size, mlp.out_dim))
test.assertTrue(torch.allclose(torch.sum(yhat0, dim=1), torch.tensor(1.0)))
test.assertIsNotNone(yhat0.grad_fn)

yhat0
/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/torch/nn/modules/module.py:1110: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
  return forward_call(*input, **kwargs)
Out[40]:
tensor([[0.0162, 0.0133, 0.0147,  ..., 0.0156, 0.0175, 0.0171],
        [0.0165, 0.0145, 0.0147,  ..., 0.0143, 0.0161, 0.0189],
        [0.0166, 0.0146, 0.0150,  ..., 0.0136, 0.0153, 0.0200],
        ...,
        [0.0161, 0.0133, 0.0145,  ..., 0.0158, 0.0180, 0.0167],
        [0.0172, 0.0148, 0.0150,  ..., 0.0128, 0.0145, 0.0208],
        [0.0166, 0.0146, 0.0147,  ..., 0.0140, 0.0158, 0.0193]],
       grad_fn=<SoftmaxBackward0>)

MLP for Binary Classification¶

The MLP model we've implemented, while useful, is very general. For the task of binary classification, we would like to add some additional functionality to it: the ability to output a normalized score for a sample being in class one (which we interpret as a probability) and a prediction based on some threshold of this probability. In addition, we need some way to calculate a meaningful threshold based on the data and a trained model at hand.

In order to maintain generality, we'll add this functionlity in the form of a wrapper: A BinaryClassifier class that can wrap any model producing two output features, and provide the the functionality stated above.

TODO: In the hw2/classifier.py module, implement the BinaryClassifier and the missing parts of its base class, Classifier. Read the method documentation carefully and implement accordingly. You can ignore the roc_threshold method at this stage.

In [41]:
from hw2.classifier import BinaryClassifier

bmlp4 = BinaryClassifier(
    model=MLP(in_dim=2, dims=[*[10]*3, 2], nonlins=[*['relu']*3, 'none']),
    threshold=0.5
)
print(bmlp4)

# Test model
test.assertEqual(len([*bmlp4.parameters()]), 8)
test.assertIsNotNone(bmlp4(x0).grad_fn)

# Test forward
yhat0_scores = bmlp4(x0)
test.assertEqual(yhat0_scores.shape, (batch_size, 2))
test.assertFalse(torch.allclose(torch.sum(yhat0_scores, dim=1), torch.tensor(1.0)))

# Test predict_proba
yhat0_proba = bmlp4.predict_proba(x0)
test.assertEqual(yhat0_proba.shape, (batch_size, 2))
test.assertTrue(torch.allclose(torch.sum(yhat0_proba, dim=1), torch.tensor(1.0)))

# Test classify
yhat0 = bmlp4.classify(x0)
test.assertEqual(yhat0.shape, (batch_size,))
test.assertEqual(yhat0.dtype, torch.int)
test.assertTrue(all(yh_ in (0, 1) for yh_ in yhat0))
BinaryClassifier(
  (model): MLP(
    (seq): Sequential(
      (0): Linear(in_features=2, out_features=10, bias=True)
      (1): ReLU()
      (2): Linear(in_features=10, out_features=10, bias=True)
      (3): ReLU()
      (4): Linear(in_features=10, out_features=10, bias=True)
      (5): ReLU()
      (6): Linear(in_features=10, out_features=2, bias=True)
      (7): Identity()
    )
  )
  (softmax): Softmax(dim=1)
)

Training¶

Now that we have a classifier, we need to train it. We will abstract the various aspects of training such as mlutiple epochs, iterating over batches, early stopping and saving model checkpoints, into a Trainer that will take care of these concerns.

The Trainer class splits the task of training (and evaluating) models into three conceptual levels,

  • Multiple epochs - the fit method, which returns a FitResult containing losses and accuracies for all epochs.
  • Single epoch - the train_epoch and test_epoch methods, which return an EpochResult containing losses per batch and the single accuracy result of the epoch.
  • Single batch - the train_batch and test_batch methods, which return a BatchResult containing a single loss and the number of correctly classified samples in the batch.

It implements the first two levels. Inheriting classes are expected to implement the single-batch level methods since these are model and/or task specific.

TODO:

  1. Implement the Trainer's fit method and the ClassifierTrainer's train_batch/test_batch methods, in the hw2/training.py module. You may ignore the Optional parts about early stopping an model checkpoints at this stage.

  2. Set the model's architecture hyper-parameters and the optimizer hyperparameters in part3_arch_hp() and part3_optim_hp(), respectively, in hw2/answers.py.

Since this is a toy dataset, you should be able to quickly get above 85% accuracy even on the test set.

In [42]:
from hw2.training import ClassifierTrainer
from hw2.answers import part3_arch_hp, part3_optim_hp

torch.manual_seed(seed)

hp_arch = part3_arch_hp()
hp_optim = part3_optim_hp()

model = BinaryClassifier(
    model=MLP(
        in_dim=2,
        dims=[*[hp_arch['hidden_dims'],]*hp_arch['n_layers'], 2],
        nonlins=[*[hp_arch['activation'],]*hp_arch['n_layers'], hp_arch['out_activation']]
    ),
    threshold=0.5,
)
print(model)

loss_fn = hp_optim.pop('loss_fn')
optimizer = torch.optim.SGD(params=model.parameters(), **hp_optim)
trainer = ClassifierTrainer(model, loss_fn, optimizer)

fit_result = trainer.fit(dl_train, dl_valid, num_epochs=20, print_every=10);

test.assertGreaterEqual(fit_result.train_acc[-1], 85.0)
test.assertGreaterEqual(fit_result.test_acc[-1], 75.0)
BinaryClassifier(
  (model): MLP(
    (seq): Sequential(
      (0): Linear(in_features=2, out_features=16, bias=True)
      (1): ReLU()
      (2): Linear(in_features=16, out_features=16, bias=True)
      (3): ReLU()
      (4): Linear(in_features=16, out_features=16, bias=True)
      (5): ReLU()
      (6): Linear(in_features=16, out_features=16, bias=True)
      (7): ReLU()
      (8): Linear(in_features=16, out_features=2, bias=True)
      (9): Softmax(dim=None)
    )
  )
  (softmax): Softmax(dim=1)
)
--- EPOCH 1/20 ---
train_batch:   0%|          | 0/167 [00:00<?, ?it/s]
/Users/elienedjar/Library/Mobile Documents/com~apple~CloudDocs/technion/Deep Learning/hw2/hw2/classifier.py:39: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
  z = self.model.forward(x)
test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
train_batch:   0%|          | 0/167 [00:00<?, ?it/s]
test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
train_batch:   0%|          | 0/167 [00:00<?, ?it/s]
test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
train_batch:   0%|          | 0/167 [00:00<?, ?it/s]
test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
IOPub message rate exceeded.
The notebook server will temporarily stop sending output
to the client in order to avoid crashing it.
To change this limit, set the config variable
`--NotebookApp.iopub_msg_rate_limit`.

Current values:
NotebookApp.iopub_msg_rate_limit=1000.0 (msgs/sec)
NotebookApp.rate_limit_window=3.0 (secs)

test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
train_batch:   0%|          | 0/167 [00:00<?, ?it/s]
test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
train_batch:   0%|          | 0/167 [00:00<?, ?it/s]
test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
train_batch:   0%|          | 0/167 [00:00<?, ?it/s]
test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
--- EPOCH 11/20 ---
train_batch:   0%|          | 0/167 [00:00<?, ?it/s]
IOPub message rate exceeded.
The notebook server will temporarily stop sending output
to the client in order to avoid crashing it.
To change this limit, set the config variable
`--NotebookApp.iopub_msg_rate_limit`.

Current values:
NotebookApp.iopub_msg_rate_limit=1000.0 (msgs/sec)
NotebookApp.rate_limit_window=3.0 (secs)

test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
train_batch:   0%|          | 0/167 [00:00<?, ?it/s]
test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
train_batch:   0%|          | 0/167 [00:00<?, ?it/s]
test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
train_batch:   0%|          | 0/167 [00:00<?, ?it/s]
test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
train_batch:   0%|          | 0/167 [00:00<?, ?it/s]
IOPub message rate exceeded.
The notebook server will temporarily stop sending output
to the client in order to avoid crashing it.
To change this limit, set the config variable
`--NotebookApp.iopub_msg_rate_limit`.

Current values:
NotebookApp.iopub_msg_rate_limit=1000.0 (msgs/sec)
NotebookApp.rate_limit_window=3.0 (secs)

test_batch:   0%|          | 0/84 [00:00<?, ?it/s]
In [43]:
from cs236781.plot import plot_fit

plot_fit(fit_result, log_loss=False, train_test_overlay=True);

Decision Boundary¶

An important part of understanding what a non-linear classifier like our MLP is doing is visualizing it's decision boundaries. When we only have two input features, these are relatively simple to visualize, since we can simply plot our data on the plane, and evaluate our classifier on a constant 2D grid in order to approximate the decision boundary.

TODO: Implement the plot_decision_boundary_2d function in the hw2/classifier.py module.

In [44]:
from hw2.classifier import plot_decision_boundary_2d

fig, ax = plot_decision_boundary_2d(model, *dl_valid.dataset.tensors)
/Users/elienedjar/Library/Mobile Documents/com~apple~CloudDocs/technion/Deep Learning/hw2/hw2/classifier.py:39: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
  z = self.model.forward(x)

Threshold Selection¶

Another important component, especially in the context of binary classification is threshold selection. Until now, we arbitrarily chose a threshold of 0.5 when deciding the class label based on the probability score we calculated via softmax. In other words, we classified a sample to class 1 (the 'positive' class) when it's probability score was greater or equal to 0.5.

However, in real-world classifiction problems we'll need to choose our threshold wisely based on the domain-specific requirements of the problem. For example, depending on our application, we might care more about high sensitivity (correctly classifying positive examples), while for other applications specificity (correctly classifying negative examples) is more important.

One way to understand the mistakes a model is making is to look at its Confusion Matrix. From it, we easily see e.g. the false-negative rate (FNR) and false-positive rate (FPR).

Let's look at the confusion matrices on the test and validation data using the model we trained above.

In [45]:
from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay

def plot_confusion(classifier, x: np.ndarray, y: np.ndarray, ax=None):
    y_hat = classifier.classify(torch.from_numpy(x).to(torch.float32)).numpy()
    conf_mat = confusion_matrix(y, y_hat, normalize='all')
    ConfusionMatrixDisplay(conf_mat).plot(ax=ax, colorbar=False)
    
model.threshold = 0.5

_, axes = plt.subplots(1, 2, figsize=(10, 5))
axes[0].set_title("Train"); axes[1].set_title("Validation");
plot_confusion(model, X_train, y_train, ax=axes[0])
plot_confusion(model, X_valid, y_valid, ax=axes[1])
/Users/elienedjar/Library/Mobile Documents/com~apple~CloudDocs/technion/Deep Learning/hw2/hw2/classifier.py:39: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
  z = self.model.forward(x)

We can see that the model makes a different number of false-posiive and false-negative errors. Clearly, this proportion would change if the classification threshold was different.

A very common way to select the classification threshold is to find a threshold which optimally balances between the FPR and FNR. This can be done by plotting the model's ROC curve, which shows 1-FNR vs. FPR for multiple threshold values, and selecting the point closest to the ideal point ((0, 1)).

TODO: Implement the select_roc_thresh function in the hw2.classifier module.

In [46]:
from hw2.classifier import select_roc_thresh


optimal_thresh = select_roc_thresh(model, *dl_valid.dataset.tensors, plot=True)
/Users/elienedjar/Library/Mobile Documents/com~apple~CloudDocs/technion/Deep Learning/hw2/hw2/classifier.py:39: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
  z = self.model.forward(x)

Let's see the effect of our threshold selection on the confusion matrix and decision boundary.

In [47]:
model.threshold = optimal_thresh

_, axes = plt.subplots(1, 2, figsize=(10, 5))
axes[0].set_title("Train"); axes[1].set_title("Validation");
plot_confusion(model, X_train, y_train, ax=axes[0])
plot_confusion(model, X_valid, y_valid, ax=axes[1])
fig, ax = plot_decision_boundary_2d(model, *dl_valid.dataset.tensors)
/Users/elienedjar/Library/Mobile Documents/com~apple~CloudDocs/technion/Deep Learning/hw2/hw2/classifier.py:39: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
  z = self.model.forward(x)
/Users/elienedjar/Library/Mobile Documents/com~apple~CloudDocs/technion/Deep Learning/hw2/hw2/classifier.py:39: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
  z = self.model.forward(x)
/Users/elienedjar/Library/Mobile Documents/com~apple~CloudDocs/technion/Deep Learning/hw2/hw2/classifier.py:39: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
  z = self.model.forward(x)

Architecture Experiments¶

Now, equipped with the tools we've implemented so far we'll expertiment with various MLP architectures. We'll seek to study the effect of the models depth (number of hidden layers) and width (number of neurons per hidden layer) on the its decision boundaries and the resulting performance. After training, we will use the validation set for threshold selection, and seek to maximize the performance on the test set.

TODO: Implement the mlp_experiment function in hw2/experiments.py. You are free to configure any model and optimization hyperparameters however you like, except for the specified width and depth. Experiment with various options for these other hyperparameters and try to obtain the best results you can.

In [48]:
from itertools import product
from tqdm.auto import tqdm
from hw2.experiments import mlp_experiment

torch.manual_seed(seed)

depths = [1, 2, 4]
widths = [2, 8, 32, 128]
exp_configs = product(enumerate(widths), enumerate(depths))
fig, axes = plt.subplots(len(widths), len(depths), figsize=(10*len(depths), 10*len(widths)), squeeze=False)
test_accs = []

for (i, width), (j, depth) in tqdm(list(exp_configs)):
    model, thresh, valid_acc, test_acc = mlp_experiment(
        depth, width, dl_train, dl_valid, dl_test, n_epochs=10
    )
    test_accs.append(test_acc)
    fig, ax = plot_decision_boundary_2d(model, *dl_test.dataset.tensors, ax=axes[i, j])
    ax.set_title(f"{depth}, {width}")
    ax.text(ax.get_xlim()[0]*.95, ax.get_ylim()[1]*.95, f"{thresh:.2f}\n{valid_acc:.1f}%\n{test_acc:.1f}%", va="top")
    
# Assert minimal performance requirements.
# You should be able to do better than these by at least 5%.
test.assertGreaterEqual(np.min(test_accs), 75.0)
test.assertGreaterEqual(np.quantile(test_accs, 0.75), 85.0)
  0%|          | 0/12 [00:00<?, ?it/s]
with depth = 1 and width = 2 valid = 78.45519310086239, test = 83.68055820465088
with depth = 2 and width = 2 valid = 72.60967379077614, test = 83.58134627342224
with depth = 4 and width = 2 valid = 76.56542932133482, test = 83.8789701461792
with depth = 1 and width = 8 valid = 86.73415823022121, test = 88.29365372657776
with depth = 2 and width = 8 valid = 87.60404949381328, test = 89.53372836112976
with depth = 4 and width = 8 valid = 88.15148106486689, test = 89.53372836112976
with depth = 1 and width = 32 valid = 88.54518185226847, test = 90.1289701461792
with depth = 2 and width = 32 valid = 87.57780277465316, test = 87.7480149269104
with depth = 4 and width = 32 valid = 86.77540307461567, test = 88.98809552192688
with depth = 1 and width = 128 valid = 87.71653543307086, test = 89.53372836112976
with depth = 2 and width = 128 valid = 87.33783277090363, test = 90.1289701461792
with depth = 4 and width = 128 valid = 87.07911511061118, test = 89.78174328804016

Questions¶

TODO Answer the following questions. Write your answers in the appropriate variables in the module hw2/answers.py.

In [49]:
from cs236781.answers import display_answer
import hw2.answers

Question 1¶

Consider the first binary classifier you trained in this notebook and the loss/accuracy curves we plotted for it on the train and validation sets, as well as the decision boundary plot.

Based on those plots, explain qualitatively whether or now your model has:

  1. High Optimization error?
  2. High Generalization error?
  3. High Approximation error?

Explain your answers for each of the above. Since this is a qualitative question, assume "high" simply means "I would take measures in order to decrease it further".

In [50]:
display_answer(hw2.answers.part3_q1)

1 As we can see our accuracy increases along the epochs and get to more than 92%, so this translates a low optimization error. 2 As we remember our the i.i.d hypothesis isn't respected since we can observe some distance between the training set and both validation and testing set. This distance is the origin of the majority of the observed error, which mean thata big part of the loss is due to a high generalization error. 3 Finally as we can see on the decision boundary our model fit well to our samples the we assume that we have a lowapproximation error

Question 2¶

Consider the first binary classifier you trained in this notebook and the confusion matrices we plotted for it.

For the validation dataset, would you expect the FPR or the FNR to be higher, and why? Recall that you have full knowledge of the data generating process.

In [51]:
display_answer(hw2.answers.part3_q2)

Be seeing the data generation process we see that in the center of the plot of the training data we have more negative samples while in the validation and testing set we have most positive samples, which is a consequency of the rotation applied on the sets Consequently during classification we can expect to have a bigger FNR.

Question 3¶

You're training a binary classifier screening of a large cohort of patients for some disease, with the aim to detect the disease early, before any symptoms appear. You train the model on easy-to-obtain features, so screening each individual patient is simple and low-cost. In case the model classifies a patient as sick, she must then be sent to furhter testing in order to confirm the illness. Assume that these further tests are expensive and involve high-risk to the patient. Assume also that once diagnosed, a low-cost treatment exists.

You wish to screen as many people as possible at the lowest possible cost and loss of life. Would you still choose the same "optimal" point on the ROC curve as above? If not, how would you choose it? Answer these questions for two possible scenarios:

  1. A person with the disease will develop non-lethal symptoms that immediately confirm the diagnosis and can then be treated.
  2. A person with the disease shows no clear symptoms and may die with high probability if not diagnosed early enough, either by your model or by the expensive test.

Explain your answers.

In [52]:
display_answer(hw2.answers.part3_q3)

In the first we want a low False Positive rate because if a patient is diagnosed as positive while he's not he will have to make the seconds tests which are really expensive and dangerous In this case if a positive patient is diagnosed as negative, will develop his symptoms in the future and will be healed. So in that case a high false negative rate does not matter.

In the second case the we want a low False Negative rate, since if the patient is negative while being sick he will die with high probability, so we don't want to miss sick patients.

Question 4¶

Analyze your results from the Architecture Experiment.

  1. Explain the decision boundaries and model performance you obtained for the columns (fixed depth, width varies).
  2. Explain the decision boundaries and model performance you obtained for the rows (fixed width, depth varies).
  3. Compare and explain the results for the following pairs of configurations, which have the same number of total parameters:
    • depth=1, width=32 and depth=4, width=8
    • depth=1, width=128 and depth=4, width=32
  4. Explain the effect of threshold selection on the validation set: did it improve the results on the test set? why?
In [53]:
display_answer(hw2.answers.part3_q4)

1) For a fixed width, we can observe a decision boundary with very similar shape. As long as the depth increase we can observe that the decision boundary is smoother, and less look like some lines that separate the data. This smoothness allow us to get better performance on the test set.

2) For fixed depth, as the width increase we can observe that our model is able to fit more precisely to the shape of the data. The depth for which is more obvious is 2. For width of 1 it looks like our model use one line to separate the data. For width of 4 we can see that our model use 6 line to separate and for wider models the number of line used in order to separate the classes increase.

3) 1 In this case the model are very similar, because first the shapes look the same (even if we can notice smoother angles in the model with D4L8 than with D1L32). To confirm our results we can see that the train accuracies are equal and the test accuracies are almost equals. 2 In this case also the model are really similar, we can se that in both model the decision boundary separate almost the data in the same place and we can also recognize similar patterns in both model. However despite this similarity in the shape of the decision boundaries, we can observe better accuracies in the model with D=1, L = 128

4) As we can see for each experiment the test accuracy is higher than the validation accuracy (on which haven't changed the threshold). This translates that indeed the threshold selection improved the test accuracy. This is du to the fact that the validation set had undergone a similar transformation than the test set (rotation of 50° for validation and rotation of 40° for the test set)so the improvement that are made on the validation set will results to big improvements on the testing set.

In [53]:
 
$$ \newcommand{\mat}[1]{\boldsymbol {#1}} \newcommand{\mattr}[1]{\boldsymbol {#1}^\top} \newcommand{\matinv}[1]{\boldsymbol {#1}^{-1}} \newcommand{\vec}[1]{\boldsymbol {#1}} \newcommand{\vectr}[1]{\boldsymbol {#1}^\top} \newcommand{\rvar}[1]{\mathrm {#1}} \newcommand{\rvec}[1]{\boldsymbol{\mathrm{#1}}} \newcommand{\diag}{\mathop{\mathrm {diag}}} \newcommand{\set}[1]{\mathbb {#1}} \newcommand{\norm}[1]{\left\lVert#1\right\rVert} \newcommand{\pderiv}[2]{\frac{\partial #1}{\partial #2}} \newcommand{\bb}[1]{\boldsymbol{#1}} $$

Part 4: Convolutional Neural Networks¶

In this part we will explore convolution networks. We'll implement a common block-based deep CNN pattern with an without residual connections.

In [1]:
import os
import re
import sys
import glob
import numpy as np
import matplotlib.pyplot as plt
import unittest
import torch
import torchvision
import torchvision.transforms as tvtf

%matplotlib inline
%load_ext autoreload
%autoreload 2
In [2]:
seed = 42
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
plt.rcParams.update({'font.size': 12})
test = unittest.TestCase()

Convolutional layers and networks¶

Convolutional layers are the most essential building blocks of the state of the art deep learning image classification models and also play an important role in many other tasks. As we saw in the tutorial, when applied to images, convolutional layers operate on and produce volumes (3D tensors) of activations.

A convenient way to interpret convolutional layers for images is as a collection of 3D learnable filters, each of which operates on a small spatial region of the input volume. Each filter is convolved with the input volume ("slides over it"), and a dot product is computed at each location followed by a non-linearity which produces one activation. All these activations produce a 2D plane known as a feature map. Multiple feature maps (one for each filter) comprise the output volume.

A crucial property of convolutional layers is their translation equivariance, i.e. shifting the input results in and equivalently shifted output. This produces the ability to detect features regardless of their spatial location in the input.

Convolutional network architectures usually follow a pattern basic repeating blocks: one or more convolution layers, each followed by a non-linearity (generally ReLU) and then a pooling layer to reduce spatial dimensions. Usually, the number of convolutional filters increases the deeper they are in the network. These layers are meant to extract features from the input. Then, one or more fully-connected layers is used to combine the extracted features into the required number of output class scores.

Building convolutional networks with PyTorch¶

PyTorch provides all the basic building blocks needed for creating a convolutional arcitecture within the torch.nn package. Let's use them to create a basic convolutional network with the following architecture pattern:

[(CONV -> ACT)*P -> POOL]*(N/P) -> (FC -> ACT)*M -> FC

Here $N$ is the total number of convolutional layers, $P$ specifies how many convolutions to perform before each pooling layer and $M$ specifies the number of hidden fully-connected layers before the final output layer.

TODO: Complete the implementaion of the CNN class in the hw2/cnn.py module. Use PyTorch's nn.Conv2d and nn.MaxPool2d for the convolution and pooling layers. It's recommended to implement the missing functionality in the order of the class' methods.

In [3]:
from hw2.cnn import CNN

test_params = [
    dict(
        in_size=(3,100,100), out_classes=10,
        channels=[32]*4, pool_every=2, hidden_dims=[100]*2,
        conv_params=dict(kernel_size=3, stride=1, padding=1),
        activation_type='relu', activation_params=dict(),
        pooling_type='max', pooling_params=dict(kernel_size=2),
    ),
    dict(
        in_size=(3,100,100), out_classes=10,
        channels=[32]*4, pool_every=2, hidden_dims=[100]*2,
        conv_params=dict(kernel_size=5, stride=2, padding=3),
        activation_type='lrelu', activation_params=dict(negative_slope=0.05),
        pooling_type='avg', pooling_params=dict(kernel_size=3),
    ),
    dict(
        in_size=(3,100,100), out_classes=3,
        channels=[16]*5, pool_every=3, hidden_dims=[100]*1,
        conv_params=dict(kernel_size=2, stride=2, padding=2),
        activation_type='lrelu', activation_params=dict(negative_slope=0.1),
        pooling_type='max', pooling_params=dict(kernel_size=2),
    ),
]

for i, params in enumerate(test_params):
    torch.manual_seed(seed)
    net = CNN(**params)
    print(f"\n=== test {i} ===")
    print(net)

    torch.manual_seed(seed)
    test_out = net(torch.ones(1, 3, 100, 100))
    print(f'{test_out}')

    expected_out = torch.load(f'tests/assets/expected_conv_out_{i:02d}.pt')
    print(f'max_diff={torch.max(torch.abs(test_out - expected_out)).item()}')
    test.assertTrue(torch.allclose(test_out, expected_out, atol=1e-3))
Good feature extractor [32, 32, 32, 32]
YourCNN(
  (feature_extractor): Sequential(
    (0): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
    (3): Dropout(p=0.2, inplace=False)
    (4): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (5): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (6): ReLU()
    (7): Dropout(p=0.2, inplace=False)
    (8): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (9): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (10): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (11): ReLU()
    (12): Dropout(p=0.2, inplace=False)
    (13): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (14): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (15): ReLU()
    (16): Dropout(p=0.2, inplace=False)
    (17): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (mlp): MLP(
    (seq): Sequential(
      (0): Linear(in_features=20000, out_features=100, bias=True)
      (1): ReLU()
      (2): Linear(in_features=100, out_features=100, bias=True)
      (3): ReLU()
      (4): Linear(in_features=100, out_features=10, bias=True)
      (5): Identity()
    )
  )
)

=== test 0 ===
CNN(
  (feature_extractor): Sequential(
    (0): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): ReLU()
    (2): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (3): ReLU()
    (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (5): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (6): ReLU()
    (7): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (8): ReLU()
    (9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (mlp): MLP(
    (seq): Sequential(
      (0): Linear(in_features=20000, out_features=100, bias=True)
      (1): ReLU()
      (2): Linear(in_features=100, out_features=100, bias=True)
      (3): ReLU()
      (4): Linear(in_features=100, out_features=10, bias=True)
      (5): Identity()
    )
  )
)
tensor([[ 0.0745, -0.1058,  0.0928,  0.0476,  0.0057,  0.0051,  0.0938, -0.0582,
          0.0573,  0.0583]], grad_fn=<AddmmBackward0>)
max_diff=7.450580596923828e-09

=== test 1 ===
CNN(
  (feature_extractor): Sequential(
    (0): Conv2d(3, 32, kernel_size=(5, 5), stride=(2, 2), padding=(3, 3))
    (1): LeakyReLU(negative_slope=0.05)
    (2): Conv2d(32, 32, kernel_size=(5, 5), stride=(2, 2), padding=(3, 3))
    (3): LeakyReLU(negative_slope=0.05)
    (4): AvgPool2d(kernel_size=3, stride=3, padding=0)
    (5): Conv2d(32, 32, kernel_size=(5, 5), stride=(2, 2), padding=(3, 3))
    (6): LeakyReLU(negative_slope=0.05)
    (7): Conv2d(32, 32, kernel_size=(5, 5), stride=(2, 2), padding=(3, 3))
    (8): LeakyReLU(negative_slope=0.05)
    (9): AvgPool2d(kernel_size=3, stride=3, padding=0)
  )
  (mlp): MLP(
    (seq): Sequential(
      (0): Linear(in_features=32, out_features=100, bias=True)
      (1): LeakyReLU(negative_slope=0.05)
      (2): Linear(in_features=100, out_features=100, bias=True)
      (3): LeakyReLU(negative_slope=0.05)
      (4): Linear(in_features=100, out_features=10, bias=True)
      (5): Identity()
    )
  )
)
tensor([[ 0.0724, -0.0030,  0.0637, -0.0073,  0.0932, -0.0662, -0.0656,  0.0076,
          0.0193,  0.0241]], grad_fn=<AddmmBackward0>)
max_diff=1.4901161193847656e-08

=== test 2 ===
CNN(
  (feature_extractor): Sequential(
    (0): Conv2d(3, 16, kernel_size=(2, 2), stride=(2, 2), padding=(2, 2))
    (1): LeakyReLU(negative_slope=0.1)
    (2): Conv2d(16, 16, kernel_size=(2, 2), stride=(2, 2), padding=(2, 2))
    (3): LeakyReLU(negative_slope=0.1)
    (4): Conv2d(16, 16, kernel_size=(2, 2), stride=(2, 2), padding=(2, 2))
    (5): LeakyReLU(negative_slope=0.1)
    (6): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (7): Conv2d(16, 16, kernel_size=(2, 2), stride=(2, 2), padding=(2, 2))
    (8): LeakyReLU(negative_slope=0.1)
    (9): Conv2d(16, 16, kernel_size=(2, 2), stride=(2, 2), padding=(2, 2))
    (10): LeakyReLU(negative_slope=0.1)
  )
  (mlp): MLP(
    (seq): Sequential(
      (0): Linear(in_features=400, out_features=100, bias=True)
      (1): LeakyReLU(negative_slope=0.1)
      (2): Linear(in_features=100, out_features=3, bias=True)
      (3): Identity()
    )
  )
)
tensor([[-0.0004, -0.0094,  0.0817]], grad_fn=<AddmmBackward0>)
max_diff=2.7939677238464355e-09
[W NNPACK.cpp:51] Could not initialize NNPACK! Reason: Unsupported hardware.

As before, we'll wrap our model with a Classifier that provides the necessary functionality for calculating probability scores and obtaining class label predictions. This time, we'll use a simple approach that simply selects the class with the highest score.

TODO: Implement the ArgMaxClassifier in the hw2/classifier.py module.

In [4]:
from hw2.classifier import ArgMaxClassifier

model = ArgMaxClassifier(model=CNN(**test_params[0]))

test_image = torch.randint(low=0, high=256, size=(3, 100, 100), dtype=torch.float).unsqueeze(0)
test.assertEqual(model.classify(test_image).shape, (1,))
test.assertEqual(model.predict_proba(test_image).shape, (1, 10))
test.assertAlmostEqual(torch.sum(model.predict_proba(test_image)).item(), 1.0, delta=1e-3)

Let's now load CIFAR-10 to use as our dataset.

In [5]:
data_dir = os.path.expanduser('~/.pytorch-datasets')
ds_train = torchvision.datasets.CIFAR10(root=data_dir, download=True, train=True, transform=tvtf.ToTensor())
ds_test = torchvision.datasets.CIFAR10(root=data_dir, download=True, train=False, transform=tvtf.ToTensor())

print(f'Train: {len(ds_train)} samples')
print(f'Test: {len(ds_test)} samples')

x0,_ = ds_train[0]
in_size = x0.shape
num_classes = 10
print('input image size =', in_size)
Files already downloaded and verified
Files already downloaded and verified
Train: 50000 samples
Test: 10000 samples
input image size = torch.Size([3, 32, 32])

Now as usual, as a sanity test let's make sure we can overfit a tiny dataset with our model. But first we need to adapt our Trainer for PyTorch models.

TODO:

  1. Complete the implementaion of the ClassifierTrainer class in the hw2/training.py module if you haven't done so already.
  2. Set the optimizer hyperparameters in part4_optim_hp(), respectively, in hw2/answers.py.
In [6]:
from hw2.training import ClassifierTrainer
from hw2.answers import part4_optim_hp

torch.manual_seed(seed)

# Define a tiny part of the CIFAR-10 dataset to overfit it
batch_size = 2
max_batches = 25
dl_train = torch.utils.data.DataLoader(ds_train, batch_size, shuffle=False)

# Create model, loss and optimizer instances
model = ArgMaxClassifier(
    model=CNN(
        in_size, num_classes, channels=[32], pool_every=1, hidden_dims=[100],
        conv_params=dict(kernel_size=3, stride=1, padding=1),
        pooling_params=dict(kernel_size=2),
    )
)

hp_optim = part4_optim_hp()
loss_fn = hp_optim.pop('loss_fn')
optimizer = torch.optim.SGD(params=model.parameters(), **hp_optim)

# Use ClassifierTrainer to run only the training loop a few times.
trainer = ClassifierTrainer(model, loss_fn, optimizer, device)
best_acc = 0
for i in range(25):
    res = trainer.train_epoch(dl_train, max_batches=max_batches, verbose=(i%5==0))
    best_acc = res.accuracy if res.accuracy > best_acc else best_acc
    
# Test overfitting
test.assertGreaterEqual(best_acc, 90)
train_batch:   0%|          | 0/25 [00:00<?, ?it/s]
train_batch:   0%|          | 0/25 [00:00<?, ?it/s]
train_batch:   0%|          | 0/25 [00:00<?, ?it/s]
train_batch:   0%|          | 0/25 [00:00<?, ?it/s]
train_batch:   0%|          | 0/25 [00:00<?, ?it/s]

Residual Networks¶

A very common addition to the basic convolutional architecture described above are shortcut connections. First proposed by He et al. (2016), this simple addition has been shown to be crucial ingredient in order to achieve effective learning with very deep networks. Virtually all state of the art image classification models from recent years use this technique.

The idea is to add an shortcut, or skip, around every two or more convolutional layers:

This adds an easy way for the network to learn identity mappings: set the weight values to be very small. The consequence is that the convolutional layers to learn a residual mapping, i.e. some delta that is applied to the identity map, instead of actually learning a completely new mapping from scratch.

Lets start by implementing a general residual block, representing a structure similar to the above diagrams. Our residual block will be composed of:

  • A "main path" with some number of convolutional layers with ReLU between them. Optionally, we'll also apply dropout and batch normalization layers (in this order) between the convolutions, before the ReLU.
  • A "shortcut path" implementing an identity mapping around the main path. In case of a different number of input/output channels, the shortcut path should contain an additional 1x1 convolution to project the channel dimension.
  • The sum of the main and shortcut paths output is passed though a ReLU and returned.

TODO: Complete the implementation of the ResidualBlock's __init__() method in the hw2/cnn.py module.

In [7]:
from hw2.cnn import ResidualBlock

torch.manual_seed(seed)

resblock = ResidualBlock(
    in_channels=3, channels=[6, 4]*2, kernel_sizes=[3, 5]*2,
    batchnorm=True, dropout=0.2
)
print(resblock)

torch.manual_seed(seed)
test_out = resblock(torch.ones(1, 3, 32, 32))
print(f'{test_out.shape}')

expected_out = torch.load('tests/assets/expected_resblock_out.pt')
print(f'max_diff={torch.max(torch.abs(test_out - expected_out)).item()}')
test.assertTrue(torch.allclose(test_out, expected_out, atol=1e-3))
ResidualBlock(
  (main_path): Sequential(
    (0): Conv2d(3, 6, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): Dropout2d(p=0.2, inplace=False)
    (2): BatchNorm2d(6, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (3): ReLU()
    (4): Conv2d(6, 4, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
    (5): Dropout2d(p=0.2, inplace=False)
    (6): BatchNorm2d(4, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (7): ReLU()
    (8): Conv2d(4, 6, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (9): Dropout2d(p=0.2, inplace=False)
    (10): BatchNorm2d(6, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (11): ReLU()
    (12): Conv2d(6, 4, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
  )
  (shortcut_path): Sequential(
    (0): Conv2d(3, 4, kernel_size=(1, 1), stride=(1, 1), bias=False)
  )
)
torch.Size([1, 4, 32, 32])
max_diff=4.76837158203125e-07

Bottleneck Blocks¶

In the ResNet Block diagram shown above, the right block is called a bottleneck block. This type of block is mainly used deep in the network, where the feature space becomes increasingly high-dimensional (i.e. there are many channels).

Instead of applying a KxK conv layer on the original input channels, a bottleneck block first projects to a lower number of features (channels), applies the KxK conv on the result, and then projects back to the original feature space. Both projections are performed with 1x1 convolutions.

TODO: Complete the implementation of the ResidualBottleneckBlock in the hw2/cnn.py module.

In [8]:
from hw2.cnn import ResidualBottleneckBlock

torch.manual_seed(seed)
resblock_bn = ResidualBottleneckBlock(
    in_out_channels=256, inner_channels=[64, 32, 64], inner_kernel_sizes=[3, 5, 3],
    batchnorm=False, dropout=0.1, activation_type="lrelu"
)
print(resblock_bn)

# Test a forward pass
torch.manual_seed(seed)
test_in  = torch.ones(1, 256, 32, 32)
test_out = resblock_bn(test_in)
print(f'{test_out.shape}')
assert test_out.shape == test_in.shape 

expected_out = torch.load('tests/assets/expected_resblock_bn_out.pt')
print(f'max_diff={torch.max(torch.abs(test_out - expected_out)).item()}')
test.assertTrue(torch.allclose(test_out, expected_out, atol=1e-3))
ResidualBottleneckBlock(
  (main_path): Sequential(
    (0): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1))
    (1): Dropout2d(p=0.1, inplace=False)
    (2): LeakyReLU(negative_slope=0.01)
    (3): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (4): Dropout2d(p=0.1, inplace=False)
    (5): LeakyReLU(negative_slope=0.01)
    (6): Conv2d(64, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
    (7): Dropout2d(p=0.1, inplace=False)
    (8): LeakyReLU(negative_slope=0.01)
    (9): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (10): Dropout2d(p=0.1, inplace=False)
    (11): LeakyReLU(negative_slope=0.01)
    (12): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))
  )
  (shortcut_path): Sequential(
    (0): Identity()
  )
)
torch.Size([1, 256, 32, 32])
max_diff=1.1920928955078125e-07

Now, based on the ResidualBlock, we'll implement our own variation of a residual network (ResNet), with the following architecture:

[-> (CONV -> ACT)*P -> POOL]*(N/P) -> (FC -> ACT)*M -> FC
 \------- SKIP ------/

Note that $N$, $P$ and $M$ are as before, however now $P$ also controls the number of convolutional layers to add a skip-connection to.

TODO: Complete the implementation of the ResNet class in the hw2/cnn.py module. You must use your ResidualBlocks or ResidualBottleneckBlocks to group together every $P$ convolutional layers.

In [9]:
from hw2.cnn import ResNet

test_params = [
    dict(
        in_size=(3,100,100), out_classes=10, channels=[32, 64]*3,
        pool_every=4, hidden_dims=[100]*2,
        activation_type='lrelu', activation_params=dict(negative_slope=0.01),
        pooling_type='avg', pooling_params=dict(kernel_size=2),
        batchnorm=True, dropout=0.1,
        bottleneck=False
    ),
    dict(
        # create 64->16->64 bottlenecks
        in_size=(3,100,100), out_classes=5, channels=[64, 16, 64]*4,
        pool_every=3, hidden_dims=[64]*1,
        activation_type='tanh',
        pooling_type='max', pooling_params=dict(kernel_size=2),
        batchnorm=True, dropout=0.1,
        bottleneck=True
    )
]

for i, params in enumerate(test_params):
    torch.manual_seed(seed)
    net = ResNet(**params)
    print(f"\n=== test {i} ===")
    print(net)

    torch.manual_seed(seed)
    test_out = net(torch.ones(1, 3, 100, 100))
    print(f'{test_out}')
    
    expected_out = torch.load(f'tests/assets/expected_resnet_out_{i:02d}.pt')
    print(f'max_diff={torch.max(torch.abs(test_out - expected_out)).item()}')
    test.assertTrue(torch.allclose(test_out, expected_out, atol=1e-3))
=== test 0 ===
ResNet(
  (feature_extractor): Sequential(
    (0): ResidualBlock(
      (main_path): Sequential(
        (0): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (1): Dropout2d(p=0.1, inplace=False)
        (2): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (3): LeakyReLU(negative_slope=0.01)
        (4): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (5): Dropout2d(p=0.1, inplace=False)
        (6): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (7): LeakyReLU(negative_slope=0.01)
        (8): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (9): Dropout2d(p=0.1, inplace=False)
        (10): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (11): LeakyReLU(negative_slope=0.01)
        (12): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut_path): Sequential(
        (0): Conv2d(3, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      )
    )
    (1): AvgPool2d(kernel_size=2, stride=2, padding=0)
    (2): ResidualBlock(
      (main_path): Sequential(
        (0): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (1): Dropout2d(p=0.1, inplace=False)
        (2): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (3): LeakyReLU(negative_slope=0.01)
        (4): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut_path): Sequential(
        (0): Identity()
      )
    )
  )
  (mlp): MLP(
    (seq): Sequential(
      (0): Linear(in_features=160000, out_features=100, bias=True)
      (1): LeakyReLU(negative_slope=0.01)
      (2): Linear(in_features=100, out_features=100, bias=True)
      (3): LeakyReLU(negative_slope=0.01)
      (4): Linear(in_features=100, out_features=10, bias=True)
      (5): Identity()
    )
  )
)
tensor([[ 0.0422,  0.0332,  0.1870, -0.0532, -0.0742,  0.1143, -0.0617, -0.0467,
          0.0852,  0.0221]], grad_fn=<AddmmBackward0>)
max_diff=1.862645149230957e-07

=== test 1 ===
ResNet(
  (feature_extractor): Sequential(
    (0): ResidualBlock(
      (main_path): Sequential(
        (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (1): Dropout2d(p=0.1, inplace=False)
        (2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (3): Tanh()
        (4): Conv2d(64, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (5): Dropout2d(p=0.1, inplace=False)
        (6): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (7): Tanh()
        (8): Conv2d(16, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut_path): Sequential(
        (0): Conv2d(3, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      )
    )
    (1): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (2): ResidualBottleneckBlock(
      (main_path): Sequential(
        (0): Conv2d(64, 16, kernel_size=(1, 1), stride=(1, 1))
        (1): Dropout2d(p=0.1, inplace=False)
        (2): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (3): Tanh()
        (4): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (5): Dropout2d(p=0.1, inplace=False)
        (6): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (7): Tanh()
        (8): Conv2d(16, 64, kernel_size=(1, 1), stride=(1, 1))
      )
      (shortcut_path): Sequential(
        (0): Identity()
      )
    )
    (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (4): ResidualBottleneckBlock(
      (main_path): Sequential(
        (0): Conv2d(64, 16, kernel_size=(1, 1), stride=(1, 1))
        (1): Dropout2d(p=0.1, inplace=False)
        (2): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (3): Tanh()
        (4): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (5): Dropout2d(p=0.1, inplace=False)
        (6): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (7): Tanh()
        (8): Conv2d(16, 64, kernel_size=(1, 1), stride=(1, 1))
      )
      (shortcut_path): Sequential(
        (0): Identity()
      )
    )
    (5): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (6): ResidualBottleneckBlock(
      (main_path): Sequential(
        (0): Conv2d(64, 16, kernel_size=(1, 1), stride=(1, 1))
        (1): Dropout2d(p=0.1, inplace=False)
        (2): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (3): Tanh()
        (4): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (5): Dropout2d(p=0.1, inplace=False)
        (6): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (7): Tanh()
        (8): Conv2d(16, 64, kernel_size=(1, 1), stride=(1, 1))
      )
      (shortcut_path): Sequential(
        (0): Identity()
      )
    )
    (7): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (mlp): MLP(
    (seq): Sequential(
      (0): Linear(in_features=2304, out_features=64, bias=True)
      (1): Tanh()
      (2): Linear(in_features=64, out_features=5, bias=True)
      (3): Identity()
    )
  )
)
tensor([[ 0.0237, -0.1945, -0.0085, -0.4024, -0.2667]],
       grad_fn=<AddmmBackward0>)
max_diff=2.3096799850463867e-07

Questions¶

TODO Answer the following questions. Write your answers in the appropriate variables in the module hw2/answers.py.

In [10]:
from cs236781.answers import display_answer
import hw2.answers

Question 1¶

Consider the bottleneck block from the right side of the ResNet diagram above. Compare it to a regular block that performs a two 3x3 convs directly on the 256-channel input (i.e. as shown in the left side of the diagram, with a different number of channels). Explain the differences between the regular block and the bottleneck block in terms of:

  1. Number of parameters. Calculate the exact numbers for these two examples.
  2. Number of floating point operations required to compute an output (qualitative assessment).
  3. Ability to combine the input: (1) spatially (within feature maps); (2) across feature maps.
In [11]:
display_answer(hw2.answers.part4_q1)

1) Let's first compute the number of parameters of a regular block which is composed by two convolutional layers with $3\times 3 $ kernel and 256 channels as input and output, so the number of parameters of one layer is: $(3\cdot3\cdot256 + 1)\cdot256$ consequently the number of parameters in the whole block is 2\cdot((3\cdot3\cdot256 + 1)\cdot256) = 1 180 160

On the other hand the bottleneck block has 3 layers: -$(1\times 1 256 \rightarrow 64) \rightarrow (3\times 3 64 \rightarrow 64) \rightarrow (1\times 1 64 \rightarrow 256)$

So the total number of params is $(256+1)\cdot64 + (3\cdot3\cdot64 + 1)\cdot64 + (64+1)\cdot256 = 70,016$

2) Assumption: Floating point operation are $+, - , \times$ and ReLu, moreover the size of the input is $(256, H , W)$

Regular block:

Conv1: $3 \cdot 3 \cdot 256 \cdot H \cdot W \cdot 256$

Relu ($\times 2$): $2 \cdot 256 \cdot H \cdot W $

Conv2: $ 3 \cdot 3 \cdot 256 \cdot H \cdot W \cdot 256 $

Residual connection: $ 256 \cdot H \cdot W $

And the sum is: $ 1 180 416 \cdot H \cdot W$

The bottleneck block:

Conv1: $1 \cdot 1 \cdot 256 \cdot H \cdot W \cdot 64 $

Relu ($\times 2$): $2 \cdot 64 \cdot H \cdot W $

Conv2: $3 \cdot 3 \cdot 64 \cdot H \cdot W \cdot 64 $

Conv3: $1 \cdot 1 \cdot 64 \cdot H \cdot W \cdot 256 $

Residual connection: $256 \cdot H \cdot W $

Relu (256): $256 \cdot H \cdot W $

And this time the sum is $70 272 \cdot H \cdot W$

As we can see, in term of computation cost the bottleneck block is more efficient than the regular block.

3) The regular block mostly effects the input spatially, we can conclude it after seeing that in the input and the output the number of feature maps are the same

The bottleneck block effects both spatial and feature map of the input, this is due to the fact that the first convolution layer is 1X1 and consequently it's maintaining the spatial aspect of the input and map the input. After this first layer there are two layers that are similar to the Regular Block and therefore are operating on the spatial aspect of the input. So we can conclude that the bottleneck block effects mostly on the spatial aspect but also on the feature map while the Regular block effects only the input spatially.

In [11]:
 
$$ \newcommand{\mat}[1]{\boldsymbol {#1}} \newcommand{\mattr}[1]{\boldsymbol {#1}^\top} \newcommand{\matinv}[1]{\boldsymbol {#1}^{-1}} \newcommand{\vec}[1]{\boldsymbol {#1}} \newcommand{\vectr}[1]{\boldsymbol {#1}^\top} \newcommand{\rvar}[1]{\mathrm {#1}} \newcommand{\rvec}[1]{\boldsymbol{\mathrm{#1}}} \newcommand{\diag}{\mathop{\mathrm {diag}}} \newcommand{\set}[1]{\mathbb {#1}} \newcommand{\norm}[1]{\left\lVert#1\right\rVert} \newcommand{\pderiv}[2]{\frac{\partial #1}{\partial #2}} \newcommand{\bb}[1]{\boldsymbol{#1}} $$

Part 5: Convolutional Architecture Experiments¶

In this part we will explore convolution networks and the effects of their architecture on accuracy. We'll use our deep CNN implementation and perform various experiments on it while varying the architecture. Then we'll implement our own custom architecture to see whether we can get high classification results on a large subset of CIFAR-10.

Training will be performed on GPU.

In [1]:
import os
import re
import sys
import glob
import numpy as np
import matplotlib.pyplot as plt
import unittest
import torch
import torchvision
import torchvision.transforms as tvtf

%matplotlib inline
%load_ext autoreload
%autoreload 2
In [2]:
seed = 42
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
plt.rcParams.update({'font.size': 12})
test = unittest.TestCase()

Experimenting with model architectures¶

We will now perform a series of experiments that train various model configurations on a part of the CIFAR-10 dataset.

To perform the experiments, you'll need to use a machine with a GPU since training time might be too long otherwise.

Note about running on GPUs¶

Here's an example of running a forward pass on the GPU (assuming you're running this notebook on a GPU-enabled machine).

In [3]:
from hw2.cnn import ResNet

net = ResNet(
    in_size=(3,100,100), out_classes=10, channels=[32, 64]*3,
    pool_every=4, hidden_dims=[100]*2,
    pooling_type='avg', pooling_params=dict(kernel_size=2),
)
net = net.to(device)

test_image = torch.randint(low=0, high=256, size=(3, 100, 100), dtype=torch.float).unsqueeze(0)
test_image = test_image.to(device)

test_out = net(test_image)
Good feature extractor [32, 32, 32, 32]
YourCNN(
  (feature_extractor): Sequential(
    (0): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
    (3): Dropout(p=0.2, inplace=False)
    (4): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (5): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (6): ReLU()
    (7): Dropout(p=0.2, inplace=False)
    (8): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (9): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (10): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (11): ReLU()
    (12): Dropout(p=0.2, inplace=False)
    (13): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (14): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (15): ReLU()
    (16): Dropout(p=0.2, inplace=False)
    (17): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (mlp): MLP(
    (seq): Sequential(
      (0): Linear(in_features=20000, out_features=100, bias=True)
      (1): ReLU()
      (2): Linear(in_features=100, out_features=100, bias=True)
      (3): ReLU()
      (4): Linear(in_features=100, out_features=10, bias=True)
      (5): Identity()
    )
  )
)

Notice how we called .to(device) on both the model and the input tensor. Here the device is a torch.device object that we created above. If an nvidia GPU is available on the machine you're running this on, the device will be 'cuda'. When you run .to(device) on a model, it recursively goes over all the model parameter tensors and copies their memory to the GPU. Similarly, calling .to(device) on the input image also copies it.

In order to train on a GPU, you need to make sure to move all your tensors to it. You'll get errors if you try to mix CPU and GPU tensors in a computation.

In [4]:
print(f'This notebook is running with device={device}')
print(f'The model parameter tensors are also on device={next(net.parameters()).device}')
print(f'The test image is also on device={test_image.device}')
print(f'The output is therefore also on device={test_out.device}')
This notebook is running with device=cpu
The model parameter tensors are also on device=cpu
The test image is also on device=cpu
The output is therefore also on device=cpu

Notes on using course servers¶

First, please read the course servers guide carefully.

To run the experiments on the course servers, you can use the py-sbatch.sh script directly to perform a single experiment run in batch mode (since it runs python once), or use the srun command to do a single run in interactive mode. For example, running a single run of experiment 1 interactively (after conda activate of course):

srun -c 2 --gres=gpu:1 --pty python -m hw2.experiments run-exp -n test -K 32 64 -L 2 -P 2 -H 100

To perform multiple runs in batch mode with sbatch (e.g. for running all the configurations of an experiments), you can create your own script based on py-sbatch.sh and invoke whatever commands you need within it.

Don't request more than 2 CPU cores and 1 GPU device for your runs. The code won't be able to utilize more than that anyway, so you'll see no performance gain if you do. It will only cause delays for other students using the servers.

General notes for running experiments¶

  • You can run the experiments on a different machine (e.g. the course servers) and copy the results (files) to the results folder on your local machine. This notebook will only display the results, not run the actual experiment code (except for a demo run).
  • It's important to give each experiment run a name as specified by the notebook instructions later on. Each run has a run_name parameter that will also be the base name of the results file which this notebook will expect to load.
  • You will implement the code to run the experiments in the hw2/experiments.py module. This module has a CLI parser so that you can invoke it as a script and pass in all the configuration parameters for a single experiment run.
  • You should use python -m hw2.experiments run-exp to run an experiment, and not python hw2/experiments.py run-exp, regardless of how/where you run it.

Experiment 1: Network depth and number of filters¶

In this part we will test some different architecture configurations based on our CNN and ResNet. Specifically, we want to try different depths and number of features to see the effects these parameters have on the model's performance.

To do this, we'll define two extra hyperparameters for our model, K (filters_per_layer) and L (layers_per_block).

  • K is a list, containing the number of filters we want to have in our conv layers.
  • L is the number of consecutive layers with the same number of filters to use.

For example, if K=[32, 64] and L=2 it means we want two conv layers with 32 filters followed by two conv layers with 64 filters. If we also use pool_every=3, the feature-extraction part of our model will be:

Conv(X,32)->ReLu->Conv(32,32)->ReLU->Conv(32,64)->ReLU->MaxPool->Conv(64,64)->ReLU

We'll try various values of the K and L parameters in combination and see how each architecture trains. All other hyperparameters are up to you, including the choice of the optimization algorithm, the learning rate, regularization and architecture hyperparams such as pool_every and hidden_dims. Note that you should select the pool_every parameter wisely per experiment so that you don't end up with zero-width feature maps.

You can try some short manual runs to determine some good values for the hyperparameters or implement cross-validation to do it. However, the dataset size you test on should be large. If you limit the number of batches, make sure to use at least 30000 training images and 5000 validation images.

The important thing is that you state what you used, how you decided on it, and explain your results based on that.

First we need to write some code to run the experiment.

TODO:

  1. Implement the cnn_experiment() function in the hw2/experiments.py module.
  2. If you haven't done so already, it would be an excellent idea to implement the early stopping feature of the Trainer class.

The following block tests that your implementation works. It's also meant to show you that each experiment run creates a result file containing the parameters to reproduce and the FitResult object for plotting.

In [5]:
from hw2.experiments import load_experiment, cnn_experiment
from cs236781.plot import plot_fit

# Test experiment1 implementation on a few data samples and with a small model
cnn_experiment(
    'test_run', seed=seed, bs_train=50, batches=10, epochs=10, early_stopping=5,
    filters_per_layer=[32,64], layers_per_block=1, pool_every=1, hidden_dims=[100],
    model_type='resnet',
)

# There should now be a file 'test_run.json' in your `results/` folder.
# We can use it to load the results of the experiment.
cfg, fit_res = load_experiment('results/test_run_L1_K32-64.json')
_, _ = plot_fit(fit_res, train_test_overlay=True)

# And `cfg` contains the exact parameters to reproduce it
print('experiment config: ', cfg)
Files already downloaded and verified
Files already downloaded and verified
--- EPOCH 1/10 ---
[W NNPACK.cpp:51] Could not initialize NNPACK! Reason: Unsupported hardware.
train_batch:   0%|          | 0/10 [00:00<?, ?it/s]
test_batch:   0%|          | 0/10 [00:00<?, ?it/s]
--- EPOCH 2/10 ---
train_batch:   0%|          | 0/10 [00:00<?, ?it/s]
test_batch:   0%|          | 0/10 [00:00<?, ?it/s]
--- EPOCH 3/10 ---
train_batch:   0%|          | 0/10 [00:00<?, ?it/s]
test_batch:   0%|          | 0/10 [00:00<?, ?it/s]
--- EPOCH 4/10 ---
train_batch:   0%|          | 0/10 [00:00<?, ?it/s]
test_batch:   0%|          | 0/10 [00:00<?, ?it/s]
--- EPOCH 5/10 ---
train_batch:   0%|          | 0/10 [00:00<?, ?it/s]
test_batch:   0%|          | 0/10 [00:00<?, ?it/s]
--- EPOCH 6/10 ---
train_batch:   0%|          | 0/10 [00:00<?, ?it/s]
test_batch:   0%|          | 0/10 [00:00<?, ?it/s]
--- EPOCH 7/10 ---
train_batch:   0%|          | 0/10 [00:00<?, ?it/s]
test_batch:   0%|          | 0/10 [00:00<?, ?it/s]
--- EPOCH 8/10 ---
train_batch:   0%|          | 0/10 [00:00<?, ?it/s]
test_batch:   0%|          | 0/10 [00:00<?, ?it/s]
--- EPOCH 9/10 ---
train_batch:   0%|          | 0/10 [00:00<?, ?it/s]
test_batch:   0%|          | 0/10 [00:00<?, ?it/s]
--- EPOCH 10/10 ---
train_batch:   0%|          | 0/10 [00:00<?, ?it/s]
test_batch:   0%|          | 0/10 [00:00<?, ?it/s]
*** Output file ./results/test_run_L1_K32-64.json written
experiment config:  {'run_name': 'test_run', 'out_dir': './results', 'seed': 42, 'device': None, 'bs_train': 50, 'bs_test': 12, 'batches': 10, 'epochs': 10, 'early_stopping': 5, 'checkpoints': None, 'lr': 0.001, 'reg': 0.001, 'filters_per_layer': [32, 64], 'layers_per_block': 1, 'pool_every': 1, 'hidden_dims': [100], 'model_type': 'resnet', 'conv_params': {'kernel_size': 5, 'stride': 1, 'padding': 2}, 'pooling_params': {'kernel_size': 2}, 'kw': {}}

We'll use the following function to load multiple experiment results and plot them together.

In [6]:
def plot_exp_results(filename_pattern, results_dir='results'):
    fig = None
    result_files = glob.glob(os.path.join(results_dir, filename_pattern))
    result_files.sort()
    if len(result_files) == 0:
        print(f'No results found for pattern {filename_pattern}.', file=sys.stderr)
        return
    for filepath in result_files:
        m = re.match('exp\d_(\d_)?(.*)\.json', os.path.basename(filepath))
        cfg, fit_res = load_experiment(filepath)
        fig, axes = plot_fit(fit_res, fig, legend=m[2],log_loss=True)
    del cfg['filters_per_layer']
    del cfg['layers_per_block']
    print('common config: ', cfg)

Experiment 1.1: Varying the network depth (L)¶

First, we'll test the effect of the network depth on training.

Configuratons:

  • K=32 fixed, with L=2,4,8,16 varying per run
  • K=64 fixed, with L=2,4,8,16 varying per run

So 8 different runs in total.

Naming runs: Each run should be named exp1_1_L{}_K{} where the braces are placeholders for the values. For example, the first run should be named exp1_1_L2_K32.

TODO: Run the experiment on the above configuration with the CNN model. Make sure the result file names are as expected. Use the following blocks to display the results.

In [7]:
plot_exp_results('exp1_1_L*_K32*.json')
common config:  {'run_name': 'exp1_1', 'out_dir': './results', 'seed': 1949872206, 'device': None, 'bs_train': 50, 'bs_test': 12, 'batches': 100, 'epochs': 100, 'early_stopping': 5, 'checkpoints': None, 'lr': 0.001, 'reg': 0.001, 'pool_every': 5, 'hidden_dims': [100], 'model_type': 'cnn', 'kw': {}}
In [8]:
plot_exp_results('exp1_1_L*_K64*.json')
common config:  {'run_name': 'exp1_1', 'out_dir': './results', 'seed': 982957578, 'device': None, 'bs_train': 50, 'bs_test': 12, 'batches': 100, 'epochs': 100, 'early_stopping': 5, 'checkpoints': None, 'lr': 0.001, 'reg': 0.001, 'pool_every': 5, 'hidden_dims': [100], 'model_type': 'cnn', 'kw': {}}

Experiment 1.2: Varying the number of filters per layer (K)¶

Now we'll test the effect of the number of convolutional filters in each layer.

Configuratons:

  • L=2 fixed, with K=[32],[64],[128],[256] varying per run.
  • L=4 fixed, with K=[32],[64],[128],[256] varying per run.
  • L=8 fixed, with K=[32],[64],[128],[256] varying per run.

So 12 different runs in total. To clarify, each run K takes the value of a list with a single element.

Naming runs: Each run should be named exp1_2_L{}_K{} where the braces are placeholders for the values. For example, the first run should be named exp1_2_L2_K32.

TODO: Run the experiment on the above configuration with the CNN model. Make sure the result file names are as expected. Use the following blocks to display the results.

In [9]:
plot_exp_results('exp1_2_L2*.json')
common config:  {'run_name': 'exp1_2', 'out_dir': './results', 'seed': 139343871, 'device': None, 'bs_train': 50, 'bs_test': 12, 'batches': 100, 'epochs': 100, 'early_stopping': 3, 'checkpoints': None, 'lr': 0.001, 'reg': 0.001, 'pool_every': 2, 'hidden_dims': [100], 'model_type': 'cnn', 'kw': {}}
In [10]:
plot_exp_results('exp1_2_L4*.json')
common config:  {'run_name': 'exp1_2', 'out_dir': './results', 'seed': 861170647, 'device': None, 'bs_train': 50, 'bs_test': 12, 'batches': 100, 'epochs': 100, 'early_stopping': 3, 'checkpoints': None, 'lr': 0.001, 'reg': 0.001, 'pool_every': 3, 'hidden_dims': [100], 'model_type': 'cnn', 'kw': {}}
In [11]:
plot_exp_results('exp1_2_L8*.json')
common config:  {'run_name': 'exp1_2', 'out_dir': './results', 'seed': 113493996, 'device': None, 'bs_train': 50, 'bs_test': 12, 'batches': 100, 'epochs': 100, 'early_stopping': 3, 'checkpoints': None, 'lr': 0.001, 'reg': 0.001, 'pool_every': 5, 'hidden_dims': [100], 'model_type': 'cnn', 'kw': {}}

Experiment 1.3: Varying both the number of filters (K) and network depth (L)¶

Now we'll test the effect of the number of convolutional filters in each layer.

Configuratons:

  • K=[64, 128, 256] fixed with L=1,2,3,4 varying per run.

So 4 different runs in total. To clarify, each run K takes the value of an array with a three elements.

Naming runs: Each run should be named exp1_3_L{}_K{}-{}-{} where the braces are placeholders for the values. For example, the first run should be named exp1_3_L1_K64-128-256.

TODO: Run the experiment on the above configuration with the CNN model. Make sure the result file names are as expected. Use the following blocks to display the results.

In [12]:
plot_exp_results('exp1_3*.json')
common config:  {'run_name': 'exp1_3', 'out_dir': './results', 'seed': 775804031, 'device': None, 'bs_train': 50, 'bs_test': 12, 'batches': 100, 'epochs': 100, 'early_stopping': 3, 'checkpoints': None, 'lr': 0.001, 'reg': 0.001, 'pool_every': 3, 'hidden_dims': [100], 'model_type': 'cnn', 'kw': {}}

Experiment 1.4: Adding depth with Residual Networks¶

Now we'll test the effect of skip connections on the training and performance.

Configuratons:

  • K=[32] fixed with L=8,16,32 varying per run.
  • K=[64, 128, 256] fixed with L=2,4,8 varying per run.

So 6 different runs in total.

Naming runs: Each run should be named exp1_4_L{}_K{}-{}-{} where the braces are placeholders for the values.

TODO: Run the experiment on the above configuration with the ResNet model. Make sure the result file names are as expected. Use the following blocks to display the results.

In [13]:
plot_exp_results('exp1_4_L*_K32.json')
common config:  {'run_name': 'exp1_4', 'out_dir': './results', 'seed': 2000912293, 'device': None, 'bs_train': 50, 'bs_test': 12, 'batches': 100, 'epochs': 100, 'early_stopping': 3, 'checkpoints': None, 'lr': 0.001, 'reg': 0.001, 'pool_every': 5, 'hidden_dims': [100], 'model_type': 'resnet', 'kw': {}}
In [13]:
 
In [14]:
plot_exp_results('exp1_4_L*_K64*.json')
common config:  {'run_name': 'exp1_4', 'out_dir': './results', 'seed': 234810998, 'device': None, 'bs_train': 50, 'bs_test': 12, 'batches': 100, 'epochs': 100, 'early_stopping': 3, 'checkpoints': None, 'lr': 0.001, 'reg': 0.001, 'pool_every': 5, 'hidden_dims': [100], 'model_type': 'resnet', 'kw': {}}

Experiment 2: Custom network architecture¶

In this part you will create your own custom network architecture based on the CNN you've implemented.

Try to overcome some of the limitations your experiment 1 results, using what you learned in the course.

You are free to add whatever you like to the model, for instance

  • Batch normalization
  • Dropout layers
  • Skip connections, bottlenecks
  • Change kernel spatial sizes and strides
  • Custom blocks or ideas from known architectures (e.g. inception module)

Just make sure to keep the model's init API identical (or maybe just add parameters).

TODO: Implement your custom architecture in the YourCNN class within the hw2/cnn.py module.

In [15]:
from hw2.cnn import YourCNN

net = YourCNN((3,100,100), 10, channels=[32]*4, pool_every=2, hidden_dims=[100]*2)
print(net)

test_image = torch.randint(low=0, high=256, size=(3, 100, 100), dtype=torch.float).unsqueeze(0)
test_out = net(test_image)
print('out =', test_out)
Good feature extractor [32, 32, 32, 32]
YourCNN(
  (feature_extractor): Sequential(
    (0): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
    (3): Dropout(p=0.2, inplace=False)
    (4): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (5): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (6): ReLU()
    (7): Dropout(p=0.2, inplace=False)
    (8): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (9): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (10): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (11): ReLU()
    (12): Dropout(p=0.2, inplace=False)
    (13): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (14): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (15): ReLU()
    (16): Dropout(p=0.2, inplace=False)
    (17): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (mlp): MLP(
    (seq): Sequential(
      (0): Linear(in_features=20000, out_features=100, bias=True)
      (1): ReLU()
      (2): Linear(in_features=100, out_features=100, bias=True)
      (3): ReLU()
      (4): Linear(in_features=100, out_features=10, bias=True)
      (5): Identity()
    )
  )
)
out = tensor([[ 0.0614, -0.0292, -0.1989,  0.0262, -0.2623,  0.0976,  0.0347,  0.1609,
          0.2635,  0.3766]], grad_fn=<AddmmBackward0>)

Experiment 2 Configuration¶

Run your custom model on at least the following:

Configuratons:

  • K=[32, 64, 128] fixed with L=3,6,9,12 varying per run.

So 4 different runs in total. To clarify, each run K takes the value of an array with a three elements.

If you want, you can add some extra runs following the same pattern. Try to see how deep a model you can train.

Naming runs: Each run should be named exp2_L{}_K{}-{}-{}-{} where the braces are placeholders for the values. For example, the first run should be named exp2_L3_K32-64-128.

TODO: Run the experiment on the above configuration with the YourCNN model. Make sure the result file names are as expected. Use the following blocks to display the results.

In [16]:
plot_exp_results('exp2*.json')
common config:  {'run_name': 'exp2', 'out_dir': './results', 'seed': 1029746568, 'device': None, 'bs_train': 50, 'bs_test': 12, 'batches': 100, 'epochs': 40, 'early_stopping': 3, 'checkpoints': None, 'lr': 0.01, 'reg': 0.001, 'pool_every': 5, 'hidden_dims': [100], 'model_type': 'ycn', 'kw': {}}

Questions¶

TODO Answer the following questions. Write your answers in the appropriate variables in the module hw2/answers.py.

In [17]:
from cs236781.answers import display_answer
import hw2.answers

Question 1¶

Analyze your results from experiment 1.1. In particular,

  1. Explain the effect of depth on the accuracy. What depth produces the best results and why do you think that's the case?
  2. Were there values of L for which the network wasn't trainable? what causes this? Suggest two things which may be done to resolve it at least partially.
In [18]:
display_answer(hw2.answers.part5_q1)
  1. Increasing the network depth, improves accuracy both on train and on test until L=4 (for both K=32 and K=64) The best depth in this experiment is L=4. With this depth, it seems that we use the potential generalization power of this simple CNN network. For deeper network, we suffer from vanishing gradient and eventually from too small training set. For shorter network, the modeling power is weaker and we get poorer results.
  2. For L=16, the network doesn't manage to learn at all. The reason is that the gradients don't propagate through all the layers (vanishing gradients). Ways to solve that issue are to add residual blocks with shortcut connections or to add Batch Normalization layers.

Question 2¶

Analyze your results from experiment 1.2. In particular, compare to the results of experiment 1.1.

In [19]:
display_answer(hw2.answers.part5_q2)

The results differ for the several L's:

  • L = 2: K=32, 64 perform similarly and better than K=128, 256. With K=64, the network hits early stopping earlier than with K=32. K=256 performs even poorer than K=128. causes an accuracy loss
  • L = 4: for all K, the bigger is K, the better accuracy the model achieves.
  • L = 8: until K=128, the bigger is K, the better accuracy the model achieves. For K=256, it performs significantly poorer.

    Insights: it seems that there is a tradeoff between the depth and the number of kernel filers/layers. For undeep CNN and for very deep CNN, only relative small number of filters is effective (up to K=64/128) However, for mid-deep CNN, increasing the number of filter/ layer has a positive impact on accuracy.

    Comparing to 1.1: in 1.1 there are distinct accuracy scores for several depth, however in 1.2, the accuracy are closer one to each other for several K per L. For example, with L=4, the test accuracies are in a 5% range. Also, there is no model which can't learn at all like in 1,1.

    In both 1.1 and 1.2 models overfit to training set with ~10/15% accuracy gap between train and test.

Question 3¶

Analyze your results from experiment 1.3.

In [20]:
display_answer(hw2.answers.part5_q3)

We observe very stable results in terms of accuracy per L value: The larger is L, the lower are the results. Best accuracy is obtained for L=1 and worse for L=4. We can easily assume that with K=[64,128,256], the value of L is critical since for each additional L, we add 3 convolutional layers to the model. As seen before, a too deep CNN is related to vanishing gradient and lack of regularization especially with this simple ConvClassifier.

Let's observe that for L=4, the model doesn't learn at all because of the reasons cited above.

Question 4¶

Analyze your results from experiment 1.4. Compare to experiment 1.1 and 1.3.

In [21]:
display_answer(hw2.answers.part5_q4)

Here again, we get very stable results for several L values for both K=32 and K=[64,128,256]:

  • The greater is L, the poorer accuracy the model achieves.
  • The models overfit to training set for every configuration with ~20/30% difference between train and test accuracies.
  • For K=[64,128,256] and L=8, the model doesn't learn at all.

Compare with 1.1: Here again we see a clear correlation between network depth and results. Up to some depth, accuracy increases, but starting from some depth, the network is too much deep and doesn't learn from training data. The difference is that in 1.1 the accuracy gaps between them are smaller than in 1.4.

Compare with 1.3: Surprisingly, this ResNet performs approximatively the same as the CNN in 1.3. We would expect that ResNets perform better than simple CNNs in 1.3.

Question 5¶

  1. Explain your modifications to the architecture which you implemented in the YourCNN class.
  2. Analyze the results of experiment 2. Compare to experiment 1.
In [22]:
display_answer(hw2.answers.part5_q5)
  1. In YourCodeNet, we added several layers compared to ConvClassifier.

    • In feature extractor part : First we added Batch Normalization layers after each convolutional layers. Secondly, we added Dropout layer each activation. _ In classifier part: We added Dropout layer after each activation. These two augmentations increased the results on both train and test sets.
  2. The plots show that the depth L is critical in this network (given K=[32,64,128]). Indeed for L=3, we achieved accuracy of ~80% on both test and train sets. However, for bigger L, the network performs very poorly with accuracy lower than 60%. Moreover, the bigger is L, the lower is the accuracy.

    Compare to 1:

    • We see that in experiment 1, we never achieved more than 70% on both train and test, and most of the time, the model overfitted to train set. At the contrary, our model achieved similar accuracies on both train and test set at ~80%. This shows that our model addresses the problem of overfitting, which we observed in Experiment 1, by adding regularization (Dropout).
    • Moreover, in Experiment 2, we observe that early stopping is triggered later than in Experiment 1. Thus, it seems that our model offers bigger place for training.
    • Experiments 1 and 2 both show that increasing the depth hurts accuracy. It seems that the problem of vanishing gradient was not addressed by the BN layers. This may be due to the fact that skip- connections were not added in a smart way and not included at all in ConvClassifier and in our custom model.